[Hibernate-JIRA] Created: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
4:09 p.m.
DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in
multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.3.1, 3.3.0.SP1, 3.3.0.GA, 3.2.6
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
4:11 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Matthias Gommeringer commented on HHH-3608:
-------------------------------------------
Ups, the example is a bit bad formatted. I'll try it again:
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
1:07 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Adrian Shum commented on HHH-3608:
----------------------------------
I have also hit exactly the same problem in my recent project (in fact I have just
discovered it yesterday, I am coming to create issue for this but found an issue has just
been created).
An extra supportive comment for Matthias' suggested change.
One of the reason that people choose SequenceStyleGenerator is to make ID generated
synchronized with DB sequence value, and hence, we can create records directly (in DB, or
thru JDBC etc) by using the sequence value as record's primary key.
However, current implementation still has chance for causing PK collision.
For example, in case of increment size of 10, my DB sequence is set with same increment
size, and assume current value is 100
When I insert a record manually, its PK will be 100. Next value of the sequence will be
110.
If now, Hibernate get the new "hiValue" from DB, "hiValue" will be
110, and "value" will be 110- 10 = 100, which will cause PK collision once
Hibernate perform insert.
To make the change even cleaner, we can simply do
if ( hiValue < 0 || value >= hiValue) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
It will change the behavior of course. A normal upgrade of hibernate won't have any
problem. However, user should aware that if user fallback to a previous version, there may
be chance that previous PK generated by later version of Hibernate may collide with new PK
generated by former Hibnerate version
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
2:58 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Matthias Gommeringer commented on HHH-3608:
-------------------------------------------
I think the same problem exists in HiLoOptimizer when it is initialized the first time.
Obviously Adrian's version of the fix is more elegant and simpler - I've just
overseen the obvious when diving into this problem to reproduce the strange behaviour.
Hoping this is fixed soon...
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
6:18 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Hassan Hammoud commented on HHH-3608:
-------------------------------------
There might be something that I am not understanding, how do is callback.getNextValue()
called two times in this code?!! It's a if else ... Am I missing something?
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
10:51 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Adrian Shum commented on HHH-3608:
----------------------------------
For the first time the optimizer is invoked (i.e. when a generator is invoked the first
time), callback.getNextValue() is called twice, to acquire for the 'value' and
'hiValue' respectively.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
2:11 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
errorken commented on HHH-3608:
-------------------------------
[code]
[/code]
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
2:33 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
errorken commented on HHH-3608:
-------------------------------
Ok, I looked at this code and according to me this implementation is like completely
fubar, or else I'm missing something major here.
This solution will only work when there is only ONE sessionFactory active for the
table/sequence.
Multiple sessions will not trigger duplicate id's because of the proper
synchronization.
However, this solution will NOT work (and thus allow duplicate id's) when more then
one system is using the same table and sequence.
I say system, since the other system might be a JVM with hibernate (using the same
SequenceStyleGenerator) or another type of system that inserts in the same table and uses
a similar mechanism for caching id's.
First part of the problem:
When the 'value' and 'hiValue' are initialized the first time, the
sequence is called TWICE (as stated in the comments above)
See Line 173 and 181 of the nested class PooledOptimizer in OptimzeFactory.
Apparantly the second call is done to calculate the hiValue. My first thought was; why?
Why don't we simply take the increment_size ?
Like; value= getValue() and hiValue = value+incrementSize ? Anyway:
For example (A = application, T = transaction, increment size = 20) :
A1 T1 : Line 173: value = getValue() ---> sequence returns '1'
A2 T1: Line xxx : select nextVal from seq ----> sequence returns '21'
A1 T1: Line 181 : hiValue = getValue() ---> sequence returns '41'
Now hibernate will give out id's starting from 1 untill it value >= hiValue. In
other words hibernate will keep giving id's untill we reach 41.
But wait, the block starting from 21 till 41 has allready been given to application 2
=> so we will get duplicate id's starting from 21
When getValue() is called at line 181, an external system might have called the sequence
just before. There is no guarentee that the call at 181 will return a sequence value that
is sequentially after the first acquired block. Synchronization here will only prevent
this within one JVM/SessionFactory, but not with multiple systems.
Second part:
Ok, supose the initialization has happened without a problem. We were lucky and the two
sequence calls returned two sequential blocks. So value is '1' and hiValue is
'21'
At some point in time we will run out of id's and a next call to getValue() will be
done. This happens in the code at line 183.
Hower, while our first, hibernate, application was inserting, the second application did a
'select nextVal from sequence'. The sequence counter is now 41.
So, at line 183 we will retrieve the nextValue() and hiValue becomes '61'
Now we calculate the value by subtracting the increment size, making it 41 (so value = 41
and hiValue = 61)
But wait ! application2 is already using 41-61 ! => duplicate id's
Third part:
Allthough not a real problem, doing the getValue() twice will always result in one
'block' of id's getting lost.
My increment size is 20, so when my application (and my application alone) is using the
sequence I would expect the sequence to start from 21 when I restart my application.
So, I start my application for the first time and I insert 3 records. The sequence is at
its initial value, so the id's in database will become 1,2 and 3.
Next, I restart my application, and again insert 3 records. But suddenly my id starts with
41 ?
Conclusion: the solution presented by Adrian Shum seems to be the correct one. Thats the
one I was expecting to see in the PooledOptimzer in the first place.
But maybe there are special reasons while its implemented otherwise?
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
4:12 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Matthias Gommeringer commented on HHH-3608:
-------------------------------------------
@errorken: could you please vote for this issue so that it will be fixed soon? I see you
got the point to what my testcase proves :-) The first time I saw this code I also was
wondering what the strategy behind it could be...
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
4:09 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
errorken commented on HHH-3608:
-------------------------------
done
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
2:35 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Erwin Vervaet commented on HHH-3608:
------------------------------------
Note that you can easily plug in a fixed PooledOptimizer like this:
<id name="id" type="long" column="id"
access="property">
<generator class="org.hibernate.id.enhanced.SequenceStyleGenerator">
<param name="sequence_name">seq_person</param>
<param name="increment_size">20</param>
<param name="optimizer">my.FixedPooledOptimizer</param>
</generator>
</id>
Where the FixedPooledOptimizer uses the fix propozed above by Adrian Shum.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
7:52 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Michael Berkowitz commented on HHH-3608:
----------------------------------------
I have applied Adrian Shum's fix and tested it pretty thouroughly (I ran 25
simultaneous servlets that each added records to two different tables that share the same
sequence-number generator; beforehand it consistently gave me conflicts, now i've run
it many times with nary a glitch)
I hope I'm not out of line in saying that I would have thought a simple fix like this
would have made its way into trunk code by now. Does it havesome insidious bug that
hasn't yet been mentioned here?
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
1:21 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Dean Hiller commented on HHH-3608:
----------------------------------
I assume with this you have to set up the db to increment by 20 every time if you have the
optimizer set for 20 values each time? is that correct?
Is there anyway to change the sql instead so don't have to change the database to
increment by 20..ie. doing a select of 20 values from the sequence instead so that other
apps that don't use hibernate are not grabbing a nextval of 40 and then wasting 41,
42, 43 as they don't have this feature(I am trying to avoid modifying those 3 apps)
thanks,
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
3:05 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Christophe Vanfleteren commented on HHH-3608:
---------------------------------------------
@Daniel:
You are correct, but what you are proposing is a useless feature. There is very little to
gain in going to fetch a sequence value 20 times in a row vs fetching a sequence value
only when you need it.
The whole gain in having a larger increment is reducing the number of roundtrips to the
db. If you have other apps that also use the sequence, either adjust them, or don't
use a pooled configuration.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
10:38 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Dean Hiller commented on HHH-3608:
----------------------------------
who is Daniel? I am not too into sql, but was hoping there was a way to select 20 values
via sql all at once(not 20 times). Either way, we figured it out and modified the db
tables config and have our own PooledOptimizer now copied from hibernates and modified as
stated. just would be nice not to have to touch the db but I am not sure what that would
look like but I thought you could do something in oracle for that(grabbing 20 and
increasing by 20 in one shot...I would have to ask the dba again) without a config
change(though I know you can do a config change too which is what we ended up doing).
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
6:52 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Björn Hamrin commented on HHH-3608:
-----------------------------------
I've had similar problems with duplicate primary keys from an Oracle sequence. I agree
that the problem is the two consecutive calls to callback.getNextValue() and the
assumption that no other thread or VM manages to call callback.getNextValue() in between.
The other odd thing about the two calls to callback.getNextValue() in the generate()
function in PooledOptimizer is that always leaves an extra large gap in the sequence when
a server starts up, as noted by errorken (part 3). The solution proposed by Adrian Shum is
to use callback.getNextValue() as the bottom value instead of the top value. That should
certainly work but changes the behavior. I would simply propose that one always set the
value to hiValue - incrementSize, when callback.getNextValue() has been called. The only
time callback.getNextValue() needs to be called twice is when a sequence is completely
new, since it is likely to return 1 the very first time. I would would propose a change of
the generate() to the following:
public Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
hiValue = callback.getNextValue();
if ( hiValue <= incrementSize )
hiValue = callback.getNextValue();
value = hiValue - incrementSize;
}
else if ( value >= hiValue ) {
hiValue = callback.getNextValue();
value = hiValue - incrementSize;
}
return make( value++ );
}
I have tested this by plugging in a custom optimizer and running some tests. It appears to
work. If anyone can see problem with this solution I would appreciate feedback.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
9:01 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Manuel Dominguez Sarmiento commented on HHH-3608:
-------------------------------------------------
We implemented a CustomPooledOptimizer to fix the gap problem in the standard
implementation. See the attached file.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
9:01 a.m.
New subject: [Hibernate-JIRA] Updated: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Manuel Dominguez Sarmiento updated HHH-3608:
--------------------------------------------
Attachment: CustomPooledOptimizer.java
CustomPooledOptimizer fixes ID gaps produced by standard implementation. This class has
been in use in our production system for over a year, no problems reported yet.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
2:52 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
So much to cover...
Why the 2 selects? Because Hibernate uses the sequence to "reserve" its block
of id values. For the sake of simplicity lets say you have "start with 1 increment
by 10". When you first obtain the next value from the table its value is 1.
However, the values you really want Hibernate to generate are 1-11. The second select
bumps the "current value" in the sequence to 11. So its not just about
calculating the upper limit for Hibernate's use in memory. This is by far the safer
approach and as it is in use in existing applications will not be getting changed. As
noted, if you do not like this behavior it is easy enough to plugin in your own optimizer
or generator.
Unfortunately there is no way that I know of to "synchronize" on a sequence.
You can't acquire pessimistic locks on it which would be the safest and is actually
the only way to have it work in both the "other jvm" and "external
app" use cases. So, given that the 2 callback.getNextValue() calls will remain in
place, what is it you suggest?
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
3:02 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
Another thought is that we could add another optimizer. I already added one new one
today. Not sure what this one would get called. "alt-pooled"?
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
3:44 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
maybe "pooled-lo" since it is the lower bucket value that the sequence value
represents. In the current code sequence value 11 (w/ increment = 10) means Hibernate is
currently serving values 1-11. In this suggestion sequence value 11 would mean Hibernate
is currently serving value 11-21
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
3:55 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Bryan Loofbourrow commented on HHH-3608:
----------------------------------------
Steve, if the two calls must stay in place, then what about this?
- If there are two calls, something might act in between them, causing the possibility of
more than one process thinking it owns an ID
- If something does that, then that will affect the value returned by the second call
So what if the code checked to see whether secondResult - firstResult == increment. If it
did not, it would then assume concurrency issues had occurred, use secondResult for the
new "low" value, and loop back to repeat the "high" value request
process. If that approach is workable, I'd think that it would not affect existing
applications in any adverse way, so could be applied to the existing optimizer. I am not
thoroughly familiar with the issues here, so I may be missing some essential point.
With this approach, there may be an edge case in which multiple processes interleave
perfectly to produce a form of infinite-looping deadlock. So it may be better to fall back
on simply using the new "low" value as a single value, and skipping reserving
the whole increment window for that go around.That way, the worse-case interleave would
simply skip sequence-caching and make some IDs go unused, which seem like tolerable
side-effects for a rare case.
BTW if you have time, I'd find it very useful to hear an explanation of why you regard
the existing code as "safer" despite the flaw demonstrated by the reporter's
test case, and despite the evidence that a single-request case has apparently been
surviving well in both the test case failed by the existing code, and in a production
environment.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
4 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Matthias Gommeringer commented on HHH-3608:
-------------------------------------------
Hi Steve,
thanks for reading this and for your reply (finally ;-) )
I do not agree (or not understand) with your statement:
>>>
... The second select bumps the "current
value" in the sequence to 11...
<<<<
My understanding is as follows:
DB sequence starts with 1 increment by 10
Obtaining next value for the first time will bump DB sequence to 11 reserving 1-10 for
hibernate. With the second select the "current value" will be bumped to 21.
Furthermore my understanding of DB sequences is that they guarantee to generate unique
numbers for several different users.
Excerpt from
http://download-uk.oracle.com/docs/cd/B13789_01/server.101/b10759/stateme...:
<<Use the CREATE SEQUENCE statement to create a sequence, which is a database object
from which multiple users may generate unique integers.>>
So if sequences are properly used the DB guarantees that there are no duplicates. Sorry,
but I'm still convinced that this is a bug. Please correct me if I'm wrong.
But luckily a custom optimizer can be registered with hibernate :-)
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
4:06 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Ryan Klems commented on HHH-3608:
---------------------------------
Steve, I think you have a misunderstanding of how the sequence works, there is no need to
synchronize or make 2 calls to get the next value, the increment is atomic. If you have a
start 1 increment 10, the first select will get 1, next 11, next 21, etc. By getting 1
and knowing that the increment is 10, you have already reserved 1-10 for yourself. If you
make the 2nd select and get 11, you have actually just reserved 1-20, not 1-11. It
doesn't matter how many jvms or application instances are hammering on that sequence.
The current behavior of the optimizer is fundamentally broken.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
4:23 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
Ryan, I have developed database applications for nearly 10 years. I understand sequences.
So please dont use rude comments such as this. There is no "fundamentally
broken" here except for you missing the point that "By getting 1 and knowing
that the increment is 10, you have already reserved 1-10 for yourself." is merely
*an* interpretation of what that one means. As long as you and everyone else using that
sequence agree and understanding that assumption everything is good. Now what happens
when your application reads first value (1) and assumes it has reserved 1-10, but another
application reads the next value (11) and assumes it has reserved 1-10? The assumption
here is whether the database is storing the upper value or the lower. Just because *you*
think it should be interpreted as lower does not make it a universal truth.
Bryan, that might work though it would come down to how (if) the calls got interleaved.
I'll have to think on that some more. As to your question about why I think its
safer, see my point above to Ryan about different assumptions as to the sequence value.
Matthias that is completely not true on any databases supporting sequences I know of.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
4:35 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Ryan Klems commented on HHH-3608:
---------------------------------
Steve, I'm saying its fundamentally broken because if you hit the sequence twice, and
make any sort of assumption that there is any sort of cohesion between those selects _is_
fundamentally broken. It's called a race, which obviously many of us who have hit
this bug have encountered. Yes, you are correct that it was an assumption on my part that
it was getting the next values, not the previous values, however you have to make an
assumption about it one way or another... either you are getting the next set or the
previous set, and all systems using it must agree. Perhaps the most sense is to break
this into pre and post versions, but making 2 selects will only work when there is only
one thing hitting the database.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
4:57 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Drew Davidson commented on HHH-3608:
------------------------------------
Steve, I don't think Ryan was being rude at all - he was being descriptively correct
in saying "fundamentally broken" as the code has issues in the fundamental
assumptions about the two calls to the sequence's nextValue. You seem very defensive
about this, and stating your credentials is pointless - everyone can make misassumptions
or mistakes.
I think that the fact that this code is demonstrably misbehaving, as seen by multiple
independent sources, should at least make you question the fundamentals and assumptions of
the code. Perhaps you should acknowledge this instead of stubbornly refusing to see that
there is a problem or lecturing others on how much you know and they don't know.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
5:48 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Björn Hamrin commented on HHH-3608:
-----------------------------------
The value returned from the sequence can be either the top value as in the current code,
or the bottom value as suggested by Adrian. If it is the top value the special case of the
sequence returning one must be handled, because -9 to 0 is not valid. If a one is returned
simply request a new value from the sequence and throw away the one. The valid region
should always be the value from the sequence plus or minus the increment, never a region
between two values from a sequence. Sorry if I'm repeating myself.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
7:18 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
Ryan, sure I am also making an assumption; as I said any interpretation of this sequence
value is necessarily an assumption. Yes there is a slight chance of these calls being
split by another call to the sequence. That was an expectable risk I thought given that
it would only occur (1) in shared access scenarios and (2) when the generator (the
optimizer portion) is initializing its state.
Drew, I am not going to get into pointless arguments. If you have something to add to the
conversation feel free to do so. Though I do have to laugh at you calling me stubborn in
this issue considering I have already offered ways to resolve it. Perhaps I misread
"Steve, I think you have a misunderstanding of how the sequence works" as
questioning my knowledge and being rude. But anyway it is not relevant, I agree.
Bjorn, its a question of whether the value you just obtained from the sequence is the
lower boundary or upper boundary of your values. Again, its an assumption on either
side.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
7:12 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
Bryan, actually that won't work. It will just delay the inevitable. Eventually there
will be an id clash.
So what I will undertake today (since no one is jumping up and down on it, which is
unfortunate since we are doing a release today) is to add a new optimizer which assumes
the sequence value is the lo value and increments from there for the specified
increment-size before going back to the sequence (which means just one read from the
sequence up front).
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
7:51 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Björn Hamrin commented on HHH-3608:
-----------------------------------
Steve, great to hear that a fix is in the works. I do think that my modification of the
generate() function to PooledOptimizer should work without changing the assumption that
the sequence returns the bottom/low value. I have run some tests successfully on it.
Posting the code appears to have reformatted it a bit though.
Also I have studied some JBoss logs that suggest that optimizers are not just created
during startup of the server. From what I have seen optimizers are can also be created
after some error resulting in an exception. Now I could be wrong about this so please
verify yourself, but if so this means that the current code is riskier that one might
first think particularity in i clustered environment.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
8:03 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Björn Hamrin commented on HHH-3608:
-----------------------------------
"without changing the assumption that the sequence returns the bottom/low
value"
Should have been:
"without changing the assumption that the sequence returns the top/high value"
Sorry.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
8:22 a.m.
New subject: [Hibernate-JIRA] Updated: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole updated HHH-3608:
--------------------------------
Assignee: Steve Ebersole
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Assignee: Steve Ebersole
Priority: Blocker
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
8:22 a.m.
New subject: [Hibernate-JIRA] Updated: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole updated HHH-3608:
--------------------------------
Priority: Major (was: Blocker)
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Assignee: Steve Ebersole
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
8:37 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
Bjorn, I think the issue is whenever the value retrieved from the sequence is the
"start with" value, which might be zero (hsqldb's default start-with, one
(the default start-with on most dbs), or any other number (I used to regularly define
start-with to provide a gap for manual insertion of "static data").
But I think it is reasonable to require the developer to specify starts-with to account
for this even if not expecting it be used for actual creation. Then we can just assume:
{code}
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue == null ) {
value = callback.getNextValue();
...
// START CHANGE
if ( ( initialValue == -1 && value.lt( incrementSize ) ) || value.eq(
initialValue.longValue() ) ) {
// the call to obtain next-value just gave us the initialValue
hiValue = callback.getNextValue();
}
else {
hiValue = value;
value = hiValue.copy().subtract( incrementSize );
}
// END CHANGE
}
...
{code}
As for the formatting, wrap your code in {noformat}
{code}
// your code here
{code}
{noformat} tags.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Assignee: Steve Ebersole
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
Time Spent: 2h
Remaining Estimate: 0h
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
8:41 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
For completeness, what I am suggesting is:
1) The above change to the existing PooledOptimizer
2) Adding a new PooledLoValueOptimizer (short-named "pooled-lo") which makes the
opposite assumption that the current sequence value is the lower boundary of the already
reserved bucket.
3) Adding an option which says which pooled optimization to prefer (since I have to assume
this is generally going to be a blanket choice for most apps); "pooled" being
the default of course.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Assignee: Steve Ebersole
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
Time Spent: 2h
Remaining Estimate: 0h
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
8:46 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
{quote}
Also I have studied some JBoss logs that suggest that optimizers are not just created
during startup of the server. From what I have seen optimizers are can also be created
after some error resulting in an exception. Now I could be wrong about this so please
verify yourself, but if so this means that the current code is riskier that one might
first think particularity in i clustered environment.
{quote}
Bjorn, the optimizer is created when the generator it is part of is
"configured". That process happens once per generator instance. The double
call happens the first time the pooled optimizer is used, which is the first time the
generator is called. Unless the error somehow rebuilds the SessionFactory I don't see
how what you say happens happens.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Assignee: Steve Ebersole
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
Time Spent: 2h
Remaining Estimate: 0h
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
9:12 a.m.
New subject: [Hibernate-JIRA] Closed: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole closed HHH-3608.
-------------------------------
Resolution: Duplicate
See "fixed by" links
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Assignee: Steve Ebersole
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
Time Spent: 2h
Remaining Estimate: 0h
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
10:40 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Björn Hamrin commented on HHH-3608:
-----------------------------------
That code is much better, but still assumes two values retrieved from the sequence to be
consecutive. This only happens initially though so it is unlikely to a problem, at least
not for long.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Assignee: Steve Ebersole
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
Time Spent: 2h
Remaining Estimate: 0h
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
8:21 p.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Steve Ebersole commented on HHH-3608:
-------------------------------------
As I said I am ok with a *potential* for a "split read". We have minimized the
likelihood of these "split reads" by minimizing when/where the double read
occurs to a very narrow circumstance. Am I totally fine with that. If you are not, then
* use the new strategy I just introduced (HHH-5218) to interpret the sequence value as the
range lower bound, or
* use a custom optimizer that does what you want, or
* trick it by starting your sequence off at something greater-than-or-equal-to the
increment size.
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Assignee: Steve Ebersole
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
Time Spent: 2h
Remaining Estimate: 0h
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
8:45 a.m.
New subject: [Hibernate-JIRA] Commented: (HHH-3608) DB sequence numbers are not unique when using the pooled SequenceStyleGenerator in multiple JVMs with the same DB
[
http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608?page=c...
]
Strong Liu commented on HHH-3608:
---------------------------------
https://issues.jboss.org/browse/JBPAPP-5763
https://issues.jboss.org/browse/JBPAPP-5762
DB sequence numbers are not unique when using the pooled
SequenceStyleGenerator in multiple JVMs with the same DB
-----------------------------------------------------------------------------------------------------------------
Key: HHH-3608
URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-3608
Project: Hibernate Core
Issue Type: Bug
Components: core
Affects Versions: 3.2.6, 3.3.0.GA, 3.3.0.SP1, 3.3.1
Environment: Hibernate 3.2.6, Oracle (any version)
Reporter: Matthias Gommeringer
Assignee: Steve Ebersole
Attachments: CustomPooledOptimizer.java, PooledOptimizerTest.java
Time Spent: 2h
Remaining Estimate: 0h
We have several Application Servers (=JVMs) running each of them using Hibernate-Objects
with the SequenceStyleGenerator+pooled configured. In unpredictable time intervals it
happens that hibernate assigns the same ID to two completely different objects which
results in a UniqueConstraintViolation exception from the database. Here an example with a
description where hibernate fails:
DB-Sequence setup:
start=0
increment=2
PooledOptimizer.generate() with 2 threads (first assignment of hiValue/value):
JVM-1 JVM-2
value=0=callback.nextval
value=2=callback.nextval
hiValue=4=callback.nextval
hiValue=6=callback.nextval
The problem's cause is in the PooledOptimizer.generate: when it initializes
the value+hiValue for the first time it invokes callback.nextValue() twice which
may provide values that do not belong to each other. The reason is that
between the assignment of "value" and "hiValue" another JVM can
retrieve a
DB sequence value from the callback which leads to an inconsistent "value" and
"hiValue"
relation (see example above).
A fix that works for multiple JVMs would be to invoke the
"callback.getNextValue()" maximum once
per "optimizer.generate()" call:
public synchronized Serializable generate(AccessCallback callback) {
if ( hiValue < 0 ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
else if ( value >= hiValue ) {
value = callback.getNextValue();
hiValue = value + incrementSize;
}
return make(value++);
}
I attached a testcase that prooves the described problem (you can see that the IDs
"2" and "3" are assigned two times).
I would be very thankful if this problem could be fixed very soon since it is a
showstopper which
occurs very unpredictably.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators....
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
5092
days inactive
5882
days old
hibernate-issues@lists.jboss.org
40 comments
15 participants
participants (15)
-
Adrian Shum (JIRA) -
Björn Hamrin (JIRA)
-
Bryan Loofbourrow (JIRA)
-
Christophe Vanfleteren (JIRA)
-
Dean Hiller (JIRA)
-
Drew Davidson (JIRA)
-
errorken (JIRA)
-
Erwin Vervaet (JIRA)
-
Hassan Hammoud (JIRA)
-
Manuel Dominguez Sarmiento (JIRA)
-
Matthias Gommeringer (JIRA)
-
Michael Berkowitz (JIRA)
-
Ryan Klems (JIRA)
-
Steve Ebersole (JIRA)
-
Strong Liu (JIRA)