[JBoss JIRA] (ISPN-8786) Inconsistency if the WriteCommand isn't successful (triangle)
by Pedro Ruivo (JIRA)
[ https://issues.jboss.org/browse/ISPN-8786?page=com.atlassian.jira.plugin.... ]
Pedro Ruivo updated ISPN-8786:
------------------------------
Description:
A stale read can happen in a backup owner if it performs a (conditional) write follow by a read operation.
If the write operation fails, it returns immediately but it may not have its local value updated. The read will return the old value.
Causes random failures:
{noformat}
*StrongCounter.compareAnd*()
{noformat}
was:
A stale read can happen in a backup owner if it performs a [conditional] write follow by a read operation.
If the write operation fails, it returns immediately but it may not have its local value updated. The read will return the old value.
Causes random failures in {{*StrongCounter.compareAnd*()}} tests.
> Inconsistency if the WriteCommand isn't successful (triangle)
> -------------------------------------------------------------
>
> Key: ISPN-8786
> URL: https://issues.jboss.org/browse/ISPN-8786
> Project: Infinispan
> Issue Type: Bug
> Components: Core
> Reporter: Pedro Ruivo
> Assignee: Pedro Ruivo
>
> A stale read can happen in a backup owner if it performs a (conditional) write follow by a read operation.
> If the write operation fails, it returns immediately but it may not have its local value updated. The read will return the old value.
> Causes random failures:
> {noformat}
> *StrongCounter.compareAnd*()
> {noformat}
--
This message was sent by Atlassian JIRA
(v7.5.0#75005)
8 years
[JBoss JIRA] (ISPN-8786) Inconsistency if the WriteCommand isn't successful (triangle)
by Pedro Ruivo (JIRA)
[ https://issues.jboss.org/browse/ISPN-8786?page=com.atlassian.jira.plugin.... ]
Pedro Ruivo updated ISPN-8786:
------------------------------
Description:
A stale read can happen in a backup owner if it performs a [conditional] write follow by a read operation.
If the write operation fails, it returns immediately but it may not have its local value updated. The read will return the old value.
Causes random failures in {{*StrongCounter.compareAnd*()}} tests.
was:
A stale read can happen in a backup owner if it performs a [conditional] write follow by a read operation.
If the write operation fails, it returns immediately but it may not have its local value updated. The read will return the old value.
> Inconsistency if the WriteCommand isn't successful (triangle)
> -------------------------------------------------------------
>
> Key: ISPN-8786
> URL: https://issues.jboss.org/browse/ISPN-8786
> Project: Infinispan
> Issue Type: Bug
> Components: Core
> Reporter: Pedro Ruivo
> Assignee: Pedro Ruivo
>
> A stale read can happen in a backup owner if it performs a [conditional] write follow by a read operation.
> If the write operation fails, it returns immediately but it may not have its local value updated. The read will return the old value.
> Causes random failures in {{*StrongCounter.compareAnd*()}} tests.
--
This message was sent by Atlassian JIRA
(v7.5.0#75005)
8 years
[JBoss JIRA] (ISPN-8893) Stack trace of a primary request of Infinispan is lost, damaging supportability severely
by Tristan Tarrant (JIRA)
[ https://issues.jboss.org/browse/ISPN-8893?page=com.atlassian.jira.plugin.... ]
Tristan Tarrant updated ISPN-8893:
----------------------------------
Fix Version/s: 8.2.10.Final
(was: 8.2.9.Final)
> Stack trace of a primary request of Infinispan is lost, damaging supportability severely
> ----------------------------------------------------------------------------------------
>
> Key: ISPN-8893
> URL: https://issues.jboss.org/browse/ISPN-8893
> Project: Infinispan
> Issue Type: Bug
> Components: Core
> Reporter: Yeray Borges
> Assignee: Yeray Borges
> Labels: top-50-list
> Fix For: 9.2.1.Final, 8.2.10.Final
>
>
> A request is handled by XNIO's "default task-NNN" thread. If the request executes a clustering cache operation, like session replication, it waits for the backup operation on another server. And if the other server failed on the backup operation, the primary server unwraps the exception to print only the remote exception. The original context of thread "default task-NNN" is lost and a user can't even identify what the primary operation was.
> This affects supportability too much. It seems only JDG 7.1 is free from this issue (it has {{AsyncInterceptorChain}}). All other releases like EAP 7.1, RHSSO which is based on EAP 7.0 has this issue.
> See the "Forum Reference" link of a mail thread for the detail.
--
This message was sent by Atlassian JIRA
(v7.5.0#75005)
8 years
[JBoss JIRA] (ISPN-8931) Clear post notifications can cause memory issues as it holds all entries locally
by William Burns (JIRA)
William Burns created ISPN-8931:
-----------------------------------
Summary: Clear post notifications can cause memory issues as it holds all entries locally
Key: ISPN-8931
URL: https://issues.jboss.org/browse/ISPN-8931
Project: Infinispan
Issue Type: Bug
Affects Versions: 9.1.3.Final, 9.2.0.Beta1
Reporter: William Burns
Assignee: William Burns
Fix For: 9.1.4.Final, 9.2.0.Beta2
Currently clear iterates over the container and stores all entries before finally invoking clear. This can cause memory issues especially for off heap as we don't want to have all of these entries in the JVM heap at the same time.
Something like
{code}
Iterator<InternalCacheEntry<Object, Object>> iterator =
dataContainer.iterator();
while (iterator.hasNext()) {
InternalCacheEntry entry = iterator.next();
iterator.remove();
notifier.notifyCacheEntryRemoved(entry.getKey(), entry.getValue(),
entry.getMetadata(), false, context, command);
}
{code}
--
This message was sent by Atlassian JIRA
(v7.5.0#75005)
8 years