1:49 a.m.
When running AS7 in a domain, we need to be able to bind each host to
explicit numa nodes when the process is launched. We need to be able to
insert a numa command before the java command for each server process
that is launched by the HostController. The numa command needs to be
configurable for each host, i.e. we might want to bind hosts to
different numa nodes or the same numa node.
Another use case that I am aware of is running the server process under
a different user account to the HostController, e.g. prefixing the java
command with something like "sudo joe java"
Having thought about our use case and a use case where the host should
start start under a different account , a possible solution could be to
modify the AS config schema to allow a generic "launch command" for each
server entity, so that the host.xml would look something like;
<server name="server-one" group="main-server-group"
launchCommand="numactl --membind 0 --cpubind 0"/>
<server name="server-two" group="main-server-group"
auto-start="true" launchCommand="numactl --membind 1 --cpubind 1">
<!-- server-two avoids port conflicts by incrementing the
ports in
the default socket-group declared in the server-group
-->
<socket-bindings port-offset="150"/>
</server>
<server name="server-three" group="main-server-group"
auto-start="true" launchCommand="sudo joe">
<!-- server-two avoids port conflicts by incrementing the
ports in
the default socket-group declared in the server-group
-->
<socket-bindings port-offset="150"/>
</server>
If a launchCommand attribute was defined for a server, you could parse
the attribute value and add the parsed commands to the ProcessBuilder
before the java command.
The main difficulty would be validating any commands added in
launchCommand attribute, not sure how problematic this would be?
Regards,
John
--
John O'Hara
johara(a)redhat.com
JBoss, by Red Hat
Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor,
Berkshire, SI4 1TE, United Kingdom.
Registered in UK and Wales under Company Registration No. 3798903 Directors: Michael
Cunningham (USA), Charlie Peters (USA), Matt Parsons (USA) and Brendan Lane (Ireland).
2:03 a.m.
That's along the lines of what I was thinking for the "different user
account" case.
Something to think about is whether "java" should be in the
launch-command[1] value (with a default value of just "java".) If not
it's really a launch-command-prefix, and then is there a use case for
launch-command-suffix?
I think using a separate child element for this instead of an attribute
is better.
Another question is whether it makes sense to allow this to be
configured at the server-group level, with servers by default inheriting
that config. My vote is "no." KISS. If there is demand for doing it at
the server-group level, we'll here about it and can add it later.
As for validation, I doubt we can do much there. I think this would be
an advanced feature where users are going to have to be careful and test
things.
[1] Note our schema uses lower case names with "-" as a separator, not
camel case. So "launch-command" not "launchCommand".
On 8/14/12 11:49 AM, John O'Hara wrote:
When running AS7 in a domain, we need to be able to bind each host
to
explicit numa nodes when the process is launched. We need to be able to
insert a numa command before the java command for each server process
that is launched by the HostController. The numa command needs to be
configurable for each host, i.e. we might want to bind hosts to
different numa nodes or the same numa node.
Another use case that I am aware of is running the server process under
a different user account to the HostController, e.g. prefixing the java
command with something like "sudo joe java"
Having thought about our use case and a use case where the host should
start start under a different account , a possible solution could be to
modify the AS config schema to allow a generic "launch command" for each
server entity, so that the host.xml would look something like;
<server name="server-one" group="main-server-group"
launchCommand="numactl --membind 0 --cpubind 0"/>
<server name="server-two" group="main-server-group"
auto-start="true" launchCommand="numactl --membind 1 --cpubind
1">
<!-- server-two avoids port conflicts by incrementing the
ports in
the default socket-group declared in the server-group
-->
<socket-bindings port-offset="150"/>
</server>
<server name="server-three" group="main-server-group"
auto-start="true" launchCommand="sudo joe">
<!-- server-two avoids port conflicts by incrementing the
ports in
the default socket-group declared in the server-group
-->
<socket-bindings port-offset="150"/>
</server>
If a launchCommand attribute was defined for a server, you could parse
the attribute value and add the parsed commands to the ProcessBuilder
before the java command.
The main difficulty would be validating any commands added in
launchCommand attribute, not sure how problematic this would be?
Regards,
John
--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat
2:28 a.m.
I think the launch-command-prefix is probably more suitable, as
ModelCombiner already does a good job at defining which java command to run.
atm I can not think of a use case for a launch-command-suffix element
re [1] - I noticed this up after I sent the email
On 08/14/2012 06:03 PM, Brian Stansberry wrote:
That's along the lines of what I was thinking for the
"different user
account" case.
Something to think about is whether "java" should be in the
launch-command[1] value (with a default value of just "java".) If not
it's really a launch-command-prefix, and then is there a use case for
launch-command-suffix?
I think using a separate child element for this instead of an attribute
is better.
Another question is whether it makes sense to allow this to be
configured at the server-group level, with servers by default inheriting
that config. My vote is "no." KISS. If there is demand for doing it at
the server-group level, we'll here about it and can add it later.
As for validation, I doubt we can do much there. I think this would be
an advanced feature where users are going to have to be careful and test
things.
[1] Note our schema uses lower case names with "-" as a separator, not
camel case. So "launch-command" not "launchCommand".
On 8/14/12 11:49 AM, John O'Hara wrote:
> When running AS7 in a domain, we need to be able to bind each host to
> explicit numa nodes when the process is launched. We need to be able to
> insert a numa command before the java command for each server process
> that is launched by the HostController. The numa command needs to be
> configurable for each host, i.e. we might want to bind hosts to
> different numa nodes or the same numa node.
>
> Another use case that I am aware of is running the server process under
> a different user account to the HostController, e.g. prefixing the java
> command with something like "sudo joe java"
>
> Having thought about our use case and a use case where the host should
> start start under a different account , a possible solution could be to
> modify the AS config schema to allow a generic "launch command" for each
> server entity, so that the host.xml would look something like;
>
> <server name="server-one"
group="main-server-group"
> launchCommand="numactl --membind 0 --cpubind 0"/>
> <server name="server-two"
group="main-server-group"
> auto-start="true" launchCommand="numactl --membind 1 --cpubind
1">
> <!-- server-two avoids port conflicts by incrementing the
> ports in
> the default socket-group declared in the server-group
> -->
> <socket-bindings port-offset="150"/>
> </server>
> <server name="server-three"
group="main-server-group"
> auto-start="true" launchCommand="sudo joe">
> <!-- server-two avoids port conflicts by incrementing the
> ports in
> the default socket-group declared in the server-group
> -->
> <socket-bindings port-offset="150"/>
> </server>
>
> If a launchCommand attribute was defined for a server, you could parse
> the attribute value and add the parsed commands to the ProcessBuilder
> before the java command.
>
> The main difficulty would be validating any commands added in
> launchCommand attribute, not sure how problematic this would be?
>
> Regards,
>
> John
>
--
John O'Hara
johara(a)redhat.com
JBoss, by Red Hat
Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor,
Berkshire, SI4 1TE, United Kingdom.
Registered in UK and Wales under Company Registration No. 3798903 Directors: Michael
Cunningham (USA), Charlie Peters (USA), Matt Parsons (USA) and Brendan Lane (Ireland).
3:14 a.m.
Then how about this:
<launch-command prefix="numactl --membind 1 --cpubind 1"/>
If we ever find a use case for a suffix, we just add the attribute.
On 8/14/12 12:28 PM, John O'Hara wrote:
I think the launch-command-prefix is probably more suitable, as
ModelCombiner already does a good job at defining which java command to run.
atm I can not think of a use case for a launch-command-suffix element
re [1] - I noticed this up after I sent the email
On 08/14/2012 06:03 PM, Brian Stansberry wrote:
> That's along the lines of what I was thinking for the "different user
> account" case.
>
> Something to think about is whether "java" should be in the
> launch-command[1] value (with a default value of just "java".) If not
> it's really a launch-command-prefix, and then is there a use case for
> launch-command-suffix?
>
> I think using a separate child element for this instead of an attribute
> is better.
>
> Another question is whether it makes sense to allow this to be
> configured at the server-group level, with servers by default inheriting
> that config. My vote is "no." KISS. If there is demand for doing it at
> the server-group level, we'll here about it and can add it later.
>
> As for validation, I doubt we can do much there. I think this would be
> an advanced feature where users are going to have to be careful and test
> things.
>
> [1] Note our schema uses lower case names with "-" as a separator, not
> camel case. So "launch-command" not "launchCommand".
>
> On 8/14/12 11:49 AM, John O'Hara wrote:
>> When running AS7 in a domain, we need to be able to bind each host to
>> explicit numa nodes when the process is launched. We need to be able to
>> insert a numa command before the java command for each server process
>> that is launched by the HostController. The numa command needs to be
>> configurable for each host, i.e. we might want to bind hosts to
>> different numa nodes or the same numa node.
>>
>> Another use case that I am aware of is running the server process under
>> a different user account to the HostController, e.g. prefixing the java
>> command with something like "sudo joe java"
>>
>> Having thought about our use case and a use case where the host should
>> start start under a different account , a possible solution could be to
>> modify the AS config schema to allow a generic "launch command" for
each
>> server entity, so that the host.xml would look something like;
>>
>> <server name="server-one"
group="main-server-group"
>> launchCommand="numactl --membind 0 --cpubind 0"/>
>> <server name="server-two"
group="main-server-group"
>> auto-start="true" launchCommand="numactl --membind 1 --cpubind
1">
>> <!-- server-two avoids port conflicts by incrementing the
>> ports in
>> the default socket-group declared in the server-group
>> -->
>> <socket-bindings port-offset="150"/>
>> </server>
>> <server name="server-three"
group="main-server-group"
>> auto-start="true" launchCommand="sudo joe">
>> <!-- server-two avoids port conflicts by incrementing the
>> ports in
>> the default socket-group declared in the server-group
>> -->
>> <socket-bindings port-offset="150"/>
>> </server>
>>
>> If a launchCommand attribute was defined for a server, you could parse
>> the attribute value and add the parsed commands to the ProcessBuilder
>> before the java command.
>>
>> The main difficulty would be validating any commands added in
>> launchCommand attribute, not sure how problematic this would be?
>>
>> Regards,
>>
>> John
>>
>
--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat
1:38 a.m.
I have an initial attempt at pulling this together;
https://github.com/johnaoahra80/jboss-as/commits/launchCommand
A couple of things that I would be grateful for some guidance on is;
1 - I have created a PREFIX attribute for the Host Server model in the
host controller. The xml parser populates this attribute with any
commands defined in <launch-command prefix="numactl --membind 1
--cpubind 1"/>. Is this an acceptable/standard way of modifying the
model for a host server?
2 - I have imported
org.jboss.as.host.controller.HostControllerLogger.ROOT_LOGGER into
ModelCombiner.java so that I can log an INFO message if a launch command
prefix has been defined. Is there a better place to log the message?
3 - I have created a new info message in ModelCombiner.java,
serverLaunchCommandPrefix, with id = 11990. This is an arbitrary id, are
there any standards for assigning id's to messages?
John
On 08/14/2012 07:14 PM, Brian Stansberry wrote:
Then how about this:
<launch-command prefix="numactl --membind 1 --cpubind 1"/>
If we ever find a use case for a suffix, we just add the attribute.
On 8/14/12 12:28 PM, John O'Hara wrote:
> I think the launch-command-prefix is probably more suitable, as
> ModelCombiner already does a good job at defining which java command to run.
>
> atm I can not think of a use case for a launch-command-suffix element
>
> re [1] - I noticed this up after I sent the email
>
> On 08/14/2012 06:03 PM, Brian Stansberry wrote:
>> That's along the lines of what I was thinking for the "different user
>> account" case.
>>
>> Something to think about is whether "java" should be in the
>> launch-command[1] value (with a default value of just "java".) If not
>> it's really a launch-command-prefix, and then is there a use case for
>> launch-command-suffix?
>>
>> I think using a separate child element for this instead of an attribute
>> is better.
>>
>> Another question is whether it makes sense to allow this to be
>> configured at the server-group level, with servers by default inheriting
>> that config. My vote is "no." KISS. If there is demand for doing it at
>> the server-group level, we'll here about it and can add it later.
>>
>> As for validation, I doubt we can do much there. I think this would be
>> an advanced feature where users are going to have to be careful and test
>> things.
>>
>> [1] Note our schema uses lower case names with "-" as a separator, not
>> camel case. So "launch-command" not "launchCommand".
>>
>> On 8/14/12 11:49 AM, John O'Hara wrote:
>
5:16 a.m.
On 15 Aug 2012, at 17:38, John O'Hara wrote:
I have an initial attempt at pulling this together;
https://github.com/johnaoahra80/jboss-as/commits/launchCommand
A couple of things that I would be grateful for some guidance on is;
1 - I have created a PREFIX attribute for the Host Server model in the
host controller. The xml parser populates this attribute with any
commands defined in <launch-command prefix="numactl --membind 1
--cpubind 1"/>. Is this an acceptable/standard way of modifying the
model for a host server?
In your DescriptionProvider, since the write attribute handler validator allows
null and expressions (which was my advice, whether expressions are needed is up for
discussion), you would need to add
root.get(ATTRIBUTES, PREFIX, EXPRESSIONS_ALLOWED).set(true);
root.get(ATTRIBUTES, PREFIX, NILLABLE).set(true);
In the parser, again if expressions are needed you should use parsePossibleExpression()
2 - I have imported
org.jboss.as.host.controller.HostControllerLogger.ROOT_LOGGER into
ModelCombiner.java so that I can log an INFO message if a launch command
prefix has been defined. Is there a better place to log the message?
3 - I have created a new info message in ModelCombiner.java,
serverLaunchCommandPrefix, with id = 11990. This is an arbitrary id, are
there any standards for assigning id's to messages?
The message id should be sequential from the previous one
John
On 08/14/2012 07:14 PM, Brian Stansberry wrote:
> Then how about this:
>
> <launch-command prefix="numactl --membind 1 --cpubind 1"/>
>
> If we ever find a use case for a suffix, we just add the attribute.
>
> On 8/14/12 12:28 PM, John O'Hara wrote:
>> I think the launch-command-prefix is probably more suitable, as
>> ModelCombiner already does a good job at defining which java command to run.
>>
>> atm I can not think of a use case for a launch-command-suffix element
>>
>> re [1] - I noticed this up after I sent the email
>>
>> On 08/14/2012 06:03 PM, Brian Stansberry wrote:
>>> That's along the lines of what I was thinking for the "different
user
>>> account" case.
>>>
>>> Something to think about is whether "java" should be in the
>>> launch-command[1] value (with a default value of just "java".) If
not
>>> it's really a launch-command-prefix, and then is there a use case for
>>> launch-command-suffix?
>>>
>>> I think using a separate child element for this instead of an attribute
>>> is better.
>>>
>>> Another question is whether it makes sense to allow this to be
>>> configured at the server-group level, with servers by default inheriting
>>> that config. My vote is "no." KISS. If there is demand for doing it
at
>>> the server-group level, we'll here about it and can add it later.
>>>
>>> As for validation, I doubt we can do much there. I think this would be
>>> an advanced feature where users are going to have to be careful and test
>>> things.
>>>
>>> [1] Note our schema uses lower case names with "-" as a separator,
not
>>> camel case. So "launch-command" not "launchCommand".
>>>
>>> On 8/14/12 11:49 AM, John O'Hara wrote:
>>
>
_______________________________________________
jboss-as7-dev mailing list
jboss-as7-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
11:52 a.m.
On 8/15/12 3:16 PM, Kabir Khan wrote:
On 15 Aug 2012, at 17:38, John O'Hara wrote:
> I have an initial attempt at pulling this together;
>
> https://github.com/johnaoahra80/jboss-as/commits/launchCommand
>
Great!
> A couple of things that I would be grateful for some guidance on
is;
>
> 1 - I have created a PREFIX attribute for the Host Server model in the
Please use launch-command-prefix as the attribute name. When users read
the resource using the CLI, that name is more self-explanatory.
> host controller. The xml parser populates this attribute with
any
> commands defined in <launch-command prefix="numactl --membind 1
> --cpubind 1"/>. Is this an acceptable/standard way of modifying the
> model for a host server?
>
Instead of using the write-attribute operation, the parser should modify
the "add" operation for the server-config and add a
launch-command-prefix parameter.
In your DescriptionProvider, since the write attribute handler
validator allows null and expressions (which was my advice, whether expressions are needed
is up for discussion), you would need to add
root.get(ATTRIBUTES, PREFIX, EXPRESSIONS_ALLOWED).set(true);
root.get(ATTRIBUTES, PREFIX, NILLABLE).set(true);
In the parser, again if expressions are needed you should use parsePossibleExpression()
We are converting most of the coding of the resources in our management
model to use the ResourceDefinition and AttributeDefinition interfaces.
I'll try tomorrow to get a patch done that converts this SERVER_CONFIG
resource to using those interfaces. That will make doing all of this
simpler.
> 2 - I have imported
> org.jboss.as.host.controller.HostControllerLogger.ROOT_LOGGER into
> ModelCombiner.java so that I can log an INFO message if a launch command
> prefix has been defined. Is there a better place to log the message?
No, ROOT_LOGGER is the one to use; it logs all the other messages
related to the HC's activities controlling the lifecycle of servers.
>
> 3 - I have created a new info message in ModelCombiner.java,
> serverLaunchCommandPrefix, with id = 11990. This is an arbitrary id, are
> there any standards for assigning id's to messages?
The message id should be sequential from the previous one
The class javadoc for HostControllerLogger explains things a bit more.
>
> John
>
> On 08/14/2012 07:14 PM, Brian Stansberry wrote:
>> Then how about this:
>>
>> <launch-command prefix="numactl --membind 1 --cpubind 1"/>
>>
>> If we ever find a use case for a suffix, we just add the attribute.
>>
>> On 8/14/12 12:28 PM, John O'Hara wrote:
>>> I think the launch-command-prefix is probably more suitable, as
>>> ModelCombiner already does a good job at defining which java command to run.
>>>
>>> atm I can not think of a use case for a launch-command-suffix element
>>>
>>> re [1] - I noticed this up after I sent the email
>>>
>>> On 08/14/2012 06:03 PM, Brian Stansberry wrote:
>>>> That's along the lines of what I was thinking for the "different
user
>>>> account" case.
>>>>
>>>> Something to think about is whether "java" should be in the
>>>> launch-command[1] value (with a default value of just "java".)
If not
>>>> it's really a launch-command-prefix, and then is there a use case
for
>>>> launch-command-suffix?
>>>>
>>>> I think using a separate child element for this instead of an attribute
>>>> is better.
>>>>
>>>> Another question is whether it makes sense to allow this to be
>>>> configured at the server-group level, with servers by default inheriting
>>>> that config. My vote is "no." KISS. If there is demand for
doing it at
>>>> the server-group level, we'll here about it and can add it later.
>>>>
>>>> As for validation, I doubt we can do much there. I think this would be
>>>> an advanced feature where users are going to have to be careful and test
>>>> things.
>>>>
>>>> [1] Note our schema uses lower case names with "-" as a
separator, not
>>>> camel case. So "launch-command" not "launchCommand".
>>>>
>>>> On 8/14/12 11:49 AM, John O'Hara wrote:
>>>
>>
>
>
> _______________________________________________
> jboss-as7-dev mailing list
> jboss-as7-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
_______________________________________________
jboss-as7-dev mailing list
jboss-as7-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat
4:26 p.m.
Adding Tom Wells to the CC. Can you guys please keep him and his team in the loop as this
will be very useful in the Admin & Config guide eventually.
Thanks,
Misty
Misty Stanley-Jones, RHCE
Supervisor, Engineering Content Services
Red Hat Brisbane (GMT+10)
☺: misty (IRC) ✉: misty(a)redhat.com ☏: + 61 7 3514 8105
On Aug 16, 2012, at 8:22 AM, Brian Stansberry <brian.stansberry(a)redhat.com> wrote:
On 8/15/12 3:16 PM, Kabir Khan wrote:
>
> On 15 Aug 2012, at 17:38, John O'Hara wrote:
>
>> I have an initial attempt at pulling this together;
>>
>> https://github.com/johnaoahra80/jboss-as/commits/launchCommand
>>
Great!
>> A couple of things that I would be grateful for some guidance on is;
>>
>> 1 - I have created a PREFIX attribute for the Host Server model in the
Please use launch-command-prefix as the attribute name. When users read
the resource using the CLI, that name is more self-explanatory.
>> host controller. The xml parser populates this attribute with any
>> commands defined in <launch-command prefix="numactl --membind 1
>> --cpubind 1"/>. Is this an acceptable/standard way of modifying the
>> model for a host server?
>>
Instead of using the write-attribute operation, the parser should modify
the "add" operation for the server-config and add a
launch-command-prefix parameter.
> In your DescriptionProvider, since the write attribute handler validator allows null
and expressions (which was my advice, whether expressions are needed is up for
discussion), you would need to add
> root.get(ATTRIBUTES, PREFIX, EXPRESSIONS_ALLOWED).set(true);
> root.get(ATTRIBUTES, PREFIX, NILLABLE).set(true);
>
> In the parser, again if expressions are needed you should use
parsePossibleExpression()
>
We are converting most of the coding of the resources in our management
model to use the ResourceDefinition and AttributeDefinition interfaces.
I'll try tomorrow to get a patch done that converts this SERVER_CONFIG
resource to using those interfaces. That will make doing all of this
simpler.
>> 2 - I have imported
>> org.jboss.as.host.controller.HostControllerLogger.ROOT_LOGGER into
>> ModelCombiner.java so that I can log an INFO message if a launch command
>> prefix has been defined. Is there a better place to log the message?
>
No, ROOT_LOGGER is the one to use; it logs all the other messages
related to the HC's activities controlling the lifecycle of servers.
>>
>> 3 - I have created a new info message in ModelCombiner.java,
>> serverLaunchCommandPrefix, with id = 11990. This is an arbitrary id, are
>> there any standards for assigning id's to messages?
>
> The message id should be sequential from the previous one
>
The class javadoc for HostControllerLogger explains things a bit more.
>>
>> John
>>
>> On 08/14/2012 07:14 PM, Brian Stansberry wrote:
>>> Then how about this:
>>>
>>> <launch-command prefix="numactl --membind 1 --cpubind 1"/>
>>>
>>> If we ever find a use case for a suffix, we just add the attribute.
>>>
>>> On 8/14/12 12:28 PM, John O'Hara wrote:
>>>> I think the launch-command-prefix is probably more suitable, as
>>>> ModelCombiner already does a good job at defining which java command to
run.
>>>>
>>>> atm I can not think of a use case for a launch-command-suffix element
>>>>
>>>> re [1] - I noticed this up after I sent the email
>>>>
>>>> On 08/14/2012 06:03 PM, Brian Stansberry wrote:
>>>>> That's along the lines of what I was thinking for the
"different user
>>>>> account" case.
>>>>>
>>>>> Something to think about is whether "java" should be in
the
>>>>> launch-command[1] value (with a default value of just
"java".) If not
>>>>> it's really a launch-command-prefix, and then is there a use case
for
>>>>> launch-command-suffix?
>>>>>
>>>>> I think using a separate child element for this instead of an
attribute
>>>>> is better.
>>>>>
>>>>> Another question is whether it makes sense to allow this to be
>>>>> configured at the server-group level, with servers by default
inheriting
>>>>> that config. My vote is "no." KISS. If there is demand for
doing it at
>>>>> the server-group level, we'll here about it and can add it
later.
>>>>>
>>>>> As for validation, I doubt we can do much there. I think this would
be
>>>>> an advanced feature where users are going to have to be careful and
test
>>>>> things.
>>>>>
>>>>> [1] Note our schema uses lower case names with "-" as a
separator, not
>>>>> camel case. So "launch-command" not
"launchCommand".
>>>>>
>>>>> On 8/14/12 11:49 AM, John O'Hara wrote:
>>>>
>>>
>>
>>
>> _______________________________________________
>> jboss-as7-dev mailing list
>> jboss-as7-dev(a)lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>
>
> _______________________________________________
> jboss-as7-dev mailing list
> jboss-as7-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>
--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat
_______________________________________________
jboss-as7-dev mailing list
jboss-as7-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
6:04 a.m.
Sure, Misty, no problem.
John, see below...
On 8/16/12 2:26 AM, Misty Stanley-Jones wrote:
Adding Tom Wells to the CC. Can you guys please keep him and his team
in
the loop as this will be very useful in the Admin & Config guide eventually.
Thanks,
Misty
Misty Stanley-Jones, RHCE
Supervisor, Engineering Content Services
Red Hat Brisbane (GMT+10)
☺: misty (IRC) ✉: misty(a)redhat.com <mailto:misty@redhat.com>☏: +61 7
3514 8105
On Aug 16, 2012, at 8:22 AM, Brian Stansberry
<brian.stansberry(a)redhat.com <mailto:brian.stansberry@redhat.com>> wrote:
> On 8/15/12 3:16 PM, Kabir Khan wrote:
>>
>> On 15 Aug 2012, at 17:38, John O'Hara wrote:
>>
>>> I have an initial attempt at pulling this together;
>>>
>>> https://github.com/johnaoahra80/jboss-as/commits/launchCommand
>>>
>
> Great!
>
>>> A couple of things that I would be grateful for some guidance on is;
>>>
>>> 1 - I have created a PREFIX attribute for the Host Server model in the
>
> Please use launch-command-prefix as the attribute name. When users read
> the resource using the CLI, that name is more self-explanatory.
>
>>> host controller. The xml parser populates this attribute with any
>>> commands defined in <launch-command prefix="numactl --membind 1
>>> --cpubind 1"/>. Is this an acceptable/standard way of modifying the
>>> model for a host server?
>>>
>
> Instead of using the write-attribute operation, the parser should modify
> the "add" operation for the server-config and add a
> launch-command-prefix parameter.
>
>> In your DescriptionProvider, since the write attribute handler
>> validator allows null and expressions (which was my advice, whether
>> expressions are needed is up for discussion), you would need to add
>> root.get(ATTRIBUTES, PREFIX, EXPRESSIONS_ALLOWED).set(true);
>> root.get(ATTRIBUTES, PREFIX, NILLABLE).set(true);
>>
>> In the parser, again if expressions are needed you should use
>> parsePossibleExpression()
>>
>
> We are converting most of the coding of the resources in our management
> model to use the ResourceDefinition and AttributeDefinition interfaces.
> I'll try tomorrow to get a patch done that converts this SERVER_CONFIG
> resource to using those interfaces. That will make doing all of this
> simpler.
>
The conversion I mentioned ^^^ is now in master. So you would need to
rebase your branch to pick it up.
In addition to the change in master, I did a commit showing how to
integrate a new attribute into the server-config resource. This isn't in
master; I just created a topic branch and put a commit in it for your
reference:
https://github.com/bstansberry/jboss-as/commit/9e7b4c8ab2720833d2aa4cb18c...
The name of the branch is
https://github.com/bstansberry/jboss-as/tree/launch-command.
The changes in that commit would replace what you did in your branch in
ModelDescriptionConstants.java. Element.java, HostModelUtil.java,
HostServerDescription.java, HostXml.java and LocalDescriptions.properties.
Cheers,
--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat
11:10 p.m.
I have an updated version of the launch command taking into account the
suggestions made below.
https://github.com/johnaoahra80/jboss-as/commit/de881403bdd8ea20fae89eea0...
I still have a question over the log number, the sequence of numbers
that have been allocated to HostControllerLogger.java have been
exhausted (10900-10939). I am not sure what number to allocate to the
new log message.
Thanks
John
On 08/16/2012 03:52 AM, Brian Stansberry wrote:
On 8/15/12 3:16 PM, Kabir Khan wrote:
> On 15 Aug 2012, at 17:38, John O'Hara wrote:
>
>> I have an initial attempt at pulling this together;
>>
>> https://github.com/johnaoahra80/jboss-as/commits/launchCommand
>>
Great!
>> A couple of things that I would be grateful for some guidance on is;
>>
>> 1 - I have created a PREFIX attribute for the Host Server model in the
Please use launch-command-prefix as the attribute name. When users read
the resource using the CLI, that name is more self-explanatory.
>> host controller. The xml parser populates this attribute with any
>> commands defined in <launch-command prefix="numactl --membind 1
>> --cpubind 1"/>. Is this an acceptable/standard way of modifying the
>> model for a host server?
>>
Instead of using the write-attribute operation, the parser should modify
the "add" operation for the server-config and add a
launch-command-prefix parameter.
> In your DescriptionProvider, since the write attribute handler validator allows null
and expressions (which was my advice, whether expressions are needed is up for
discussion), you would need to add
> root.get(ATTRIBUTES, PREFIX, EXPRESSIONS_ALLOWED).set(true);
> root.get(ATTRIBUTES, PREFIX, NILLABLE).set(true);
>
> In the parser, again if expressions are needed you should use
parsePossibleExpression()
>
We are converting most of the coding of the resources in our management
model to use the ResourceDefinition and AttributeDefinition interfaces.
I'll try tomorrow to get a patch done that converts this SERVER_CONFIG
resource to using those interfaces. That will make doing all of this
simpler.
>> 2 - I have imported
>> org.jboss.as.host.controller.HostControllerLogger.ROOT_LOGGER into
>> ModelCombiner.java so that I can log an INFO message if a launch command
>> prefix has been defined. Is there a better place to log the message?
No, ROOT_LOGGER is the one to use; it logs all the other messages
related to the HC's activities controlling the lifecycle of servers.
>> 3 - I have created a new info message in ModelCombiner.java,
>> serverLaunchCommandPrefix, with id = 11990. This is an arbitrary id, are
>> there any standards for assigning id's to messages?
> The message id should be sequential from the previous one
>
The class javadoc for HostControllerLogger explains things a bit more.
>> John
>>
>> On 08/14/2012 07:14 PM, Brian Stansberry wrote:
>>> Then how about this:
>>>
>>> <launch-command prefix="numactl --membind 1 --cpubind 1"/>
>>>
>>> If we ever find a use case for a suffix, we just add the attribute.
>>>
>>> On 8/14/12 12:28 PM, John O'Hara wrote:
>>>> I think the launch-command-prefix is probably more suitable, as
>>>> ModelCombiner already does a good job at defining which java command to
run.
>>>>
>>>> atm I can not think of a use case for a launch-command-suffix element
>>>>
>>>> re [1] - I noticed this up after I sent the email
>>>>
>>>> On 08/14/2012 06:03 PM, Brian Stansberry wrote:
>>>>> That's along the lines of what I was thinking for the
"different user
>>>>> account" case.
>>>>>
>>>>> Something to think about is whether "java" should be in
the
>>>>> launch-command[1] value (with a default value of just
"java".) If not
>>>>> it's really a launch-command-prefix, and then is there a use case
for
>>>>> launch-command-suffix?
>>>>>
>>>>> I think using a separate child element for this instead of an
attribute
>>>>> is better.
>>>>>
>>>>> Another question is whether it makes sense to allow this to be
>>>>> configured at the server-group level, with servers by default
inheriting
>>>>> that config. My vote is "no." KISS. If there is demand for
doing it at
>>>>> the server-group level, we'll here about it and can add it
later.
>>>>>
>>>>> As for validation, I doubt we can do much there. I think this would
be
>>>>> an advanced feature where users are going to have to be careful and
test
>>>>> things.
>>>>>
>>>>> [1] Note our schema uses lower case names with "-" as a
separator, not
>>>>> camel case. So "launch-command" not
"launchCommand".
>>>>>
>>>>> On 8/14/12 11:49 AM, John O'Hara wrote:
>>
>> _______________________________________________
>> jboss-as7-dev mailing list
>> jboss-as7-dev(a)lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>
> _______________________________________________
> jboss-as7-dev mailing list
> jboss-as7-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>
--
John O'Hara
johara(a)redhat.com
JBoss, by Red Hat
Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor,
Berkshire, SI4 1TE, United Kingdom.
Registered in UK and Wales under Company Registration No. 3798903 Directors: Michael
Cunningham (USA), Charlie Peters (USA), Matt Parsons (USA) and Brendan Lane (Ireland).
6:16 p.m.
Go to http://community.jboss.org/docs/DOC-16810 and reserve a new range, and be sure to
update the javadoc in both HostControllerLogger and HostControllerMessages
On 28 Aug 2012, at 15:10, John O'Hara wrote:
I have an updated version of the launch command taking into account
the
suggestions made below.
https://github.com/johnaoahra80/jboss-as/commit/de881403bdd8ea20fae89eea0...
I still have a question over the log number, the sequence of numbers
that have been allocated to HostControllerLogger.java have been
exhausted (10900-10939). I am not sure what number to allocate to the
new log message.
Thanks
John
On 08/16/2012 03:52 AM, Brian Stansberry wrote:
> On 8/15/12 3:16 PM, Kabir Khan wrote:
>> On 15 Aug 2012, at 17:38, John O'Hara wrote:
>>
>>> I have an initial attempt at pulling this together;
>>>
>>> https://github.com/johnaoahra80/jboss-as/commits/launchCommand
>>>
> Great!
>
>>> A couple of things that I would be grateful for some guidance on is;
>>>
>>> 1 - I have created a PREFIX attribute for the Host Server model in the
> Please use launch-command-prefix as the attribute name. When users read
> the resource using the CLI, that name is more self-explanatory.
>
>>> host controller. The xml parser populates this attribute with any
>>> commands defined in <launch-command prefix="numactl --membind 1
>>> --cpubind 1"/>. Is this an acceptable/standard way of modifying the
>>> model for a host server?
>>>
> Instead of using the write-attribute operation, the parser should modify
> the "add" operation for the server-config and add a
> launch-command-prefix parameter.
>
>> In your DescriptionProvider, since the write attribute handler validator allows
null and expressions (which was my advice, whether expressions are needed is up for
discussion), you would need to add
>> root.get(ATTRIBUTES, PREFIX, EXPRESSIONS_ALLOWED).set(true);
>> root.get(ATTRIBUTES, PREFIX, NILLABLE).set(true);
>>
>> In the parser, again if expressions are needed you should use
parsePossibleExpression()
>>
> We are converting most of the coding of the resources in our management
> model to use the ResourceDefinition and AttributeDefinition interfaces.
> I'll try tomorrow to get a patch done that converts this SERVER_CONFIG
> resource to using those interfaces. That will make doing all of this
> simpler.
>
>>> 2 - I have imported
>>> org.jboss.as.host.controller.HostControllerLogger.ROOT_LOGGER into
>>> ModelCombiner.java so that I can log an INFO message if a launch command
>>> prefix has been defined. Is there a better place to log the message?
> No, ROOT_LOGGER is the one to use; it logs all the other messages
> related to the HC's activities controlling the lifecycle of servers.
>
>>> 3 - I have created a new info message in ModelCombiner.java,
>>> serverLaunchCommandPrefix, with id = 11990. This is an arbitrary id, are
>>> there any standards for assigning id's to messages?
>> The message id should be sequential from the previous one
>>
> The class javadoc for HostControllerLogger explains things a bit more.
>
>>> John
>>>
>>> On 08/14/2012 07:14 PM, Brian Stansberry wrote:
>>>> Then how about this:
>>>>
>>>> <launch-command prefix="numactl --membind 1 --cpubind
1"/>
>>>>
>>>> If we ever find a use case for a suffix, we just add the attribute.
>>>>
>>>> On 8/14/12 12:28 PM, John O'Hara wrote:
>>>>> I think the launch-command-prefix is probably more suitable, as
>>>>> ModelCombiner already does a good job at defining which java command
to run.
>>>>>
>>>>> atm I can not think of a use case for a launch-command-suffix
element
>>>>>
>>>>> re [1] - I noticed this up after I sent the email
>>>>>
>>>>> On 08/14/2012 06:03 PM, Brian Stansberry wrote:
>>>>>> That's along the lines of what I was thinking for the
"different user
>>>>>> account" case.
>>>>>>
>>>>>> Something to think about is whether "java" should be in
the
>>>>>> launch-command[1] value (with a default value of just
"java".) If not
>>>>>> it's really a launch-command-prefix, and then is there a use
case for
>>>>>> launch-command-suffix?
>>>>>>
>>>>>> I think using a separate child element for this instead of an
attribute
>>>>>> is better.
>>>>>>
>>>>>> Another question is whether it makes sense to allow this to be
>>>>>> configured at the server-group level, with servers by default
inheriting
>>>>>> that config. My vote is "no." KISS. If there is demand
for doing it at
>>>>>> the server-group level, we'll here about it and can add it
later.
>>>>>>
>>>>>> As for validation, I doubt we can do much there. I think this
would be
>>>>>> an advanced feature where users are going to have to be careful
and test
>>>>>> things.
>>>>>>
>>>>>> [1] Note our schema uses lower case names with "-" as a
separator, not
>>>>>> camel case. So "launch-command" not
"launchCommand".
>>>>>>
>>>>>> On 8/14/12 11:49 AM, John O'Hara wrote:
>>>
>>> _______________________________________________
>>> jboss-as7-dev mailing list
>>> jboss-as7-dev(a)lists.jboss.org
>>> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>>
>> _______________________________________________
>> jboss-as7-dev mailing list
>> jboss-as7-dev(a)lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>>
>
--
John O'Hara
johara(a)redhat.com
JBoss, by Red Hat
Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor,
Berkshire, SI4 1TE, United Kingdom.
Registered in UK and Wales under Company Registration No. 3798903 Directors: Michael
Cunningham (USA), Charlie Peters (USA), Matt Parsons (USA) and Brendan Lane (Ireland).
_______________________________________________
jboss-as7-dev mailing list
jboss-as7-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
10:39 p.m.
The additional range 16500-16599 has already been allocated to these
classes, but HostControllerLogger hasn't used any in that range yet. So
all that's needed is to add javadoc and comments to
HostControllerMessages saying that class will only use 16500-16549, and
then HostControllerLogger can use 16550-16599.
On 8/29/12 4:16 AM, Kabir Khan wrote:
Go to http://community.jboss.org/docs/DOC-16810 and reserve a new
range, and be sure to update the javadoc in both HostControllerLogger and
HostControllerMessages
On 28 Aug 2012, at 15:10, John O'Hara wrote:
> I have an updated version of the launch command taking into account the
> suggestions made below.
>
>
https://github.com/johnaoahra80/jboss-as/commit/de881403bdd8ea20fae89eea0...
>
> I still have a question over the log number, the sequence of numbers
> that have been allocated to HostControllerLogger.java have been
> exhausted (10900-10939). I am not sure what number to allocate to the
> new log message.
>
> Thanks
>
> John
>
> On 08/16/2012 03:52 AM, Brian Stansberry wrote:
>> On 8/15/12 3:16 PM, Kabir Khan wrote:
>>> On 15 Aug 2012, at 17:38, John O'Hara wrote:
>>>
>>>> I have an initial attempt at pulling this together;
>>>>
>>>> https://github.com/johnaoahra80/jboss-as/commits/launchCommand
>>>>
>> Great!
>>
>>>> A couple of things that I would be grateful for some guidance on is;
>>>>
>>>> 1 - I have created a PREFIX attribute for the Host Server model in the
>> Please use launch-command-prefix as the attribute name. When users read
>> the resource using the CLI, that name is more self-explanatory.
>>
>>>> host controller. The xml parser populates this attribute with any
>>>> commands defined in <launch-command prefix="numactl --membind 1
>>>> --cpubind 1"/>. Is this an acceptable/standard way of modifying
the
>>>> model for a host server?
>>>>
>> Instead of using the write-attribute operation, the parser should modify
>> the "add" operation for the server-config and add a
>> launch-command-prefix parameter.
>>
>>> In your DescriptionProvider, since the write attribute handler validator
allows null and expressions (which was my advice, whether expressions are needed is up for
discussion), you would need to add
>>> root.get(ATTRIBUTES, PREFIX, EXPRESSIONS_ALLOWED).set(true);
>>> root.get(ATTRIBUTES, PREFIX, NILLABLE).set(true);
>>>
>>> In the parser, again if expressions are needed you should use
parsePossibleExpression()
>>>
>> We are converting most of the coding of the resources in our management
>> model to use the ResourceDefinition and AttributeDefinition interfaces.
>> I'll try tomorrow to get a patch done that converts this SERVER_CONFIG
>> resource to using those interfaces. That will make doing all of this
>> simpler.
>>
>>>> 2 - I have imported
>>>> org.jboss.as.host.controller.HostControllerLogger.ROOT_LOGGER into
>>>> ModelCombiner.java so that I can log an INFO message if a launch command
>>>> prefix has been defined. Is there a better place to log the message?
>> No, ROOT_LOGGER is the one to use; it logs all the other messages
>> related to the HC's activities controlling the lifecycle of servers.
>>
>>>> 3 - I have created a new info message in ModelCombiner.java,
>>>> serverLaunchCommandPrefix, with id = 11990. This is an arbitrary id, are
>>>> there any standards for assigning id's to messages?
>>> The message id should be sequential from the previous one
>>>
>> The class javadoc for HostControllerLogger explains things a bit more.
>>
>>>> John
>>>>
>>>> On 08/14/2012 07:14 PM, Brian Stansberry wrote:
>>>>> Then how about this:
>>>>>
>>>>> <launch-command prefix="numactl --membind 1 --cpubind
1"/>
>>>>>
>>>>> If we ever find a use case for a suffix, we just add the attribute.
>>>>>
>>>>> On 8/14/12 12:28 PM, John O'Hara wrote:
>>>>>> I think the launch-command-prefix is probably more suitable, as
>>>>>> ModelCombiner already does a good job at defining which java
command to run.
>>>>>>
>>>>>> atm I can not think of a use case for a launch-command-suffix
element
>>>>>>
>>>>>> re [1] - I noticed this up after I sent the email
>>>>>>
>>>>>> On 08/14/2012 06:03 PM, Brian Stansberry wrote:
>>>>>>> That's along the lines of what I was thinking for the
"different user
>>>>>>> account" case.
>>>>>>>
>>>>>>> Something to think about is whether "java" should
be in the
>>>>>>> launch-command[1] value (with a default value of just
"java".) If not
>>>>>>> it's really a launch-command-prefix, and then is there a
use case for
>>>>>>> launch-command-suffix?
>>>>>>>
>>>>>>> I think using a separate child element for this instead of an
attribute
>>>>>>> is better.
>>>>>>>
>>>>>>> Another question is whether it makes sense to allow this to
be
>>>>>>> configured at the server-group level, with servers by default
inheriting
>>>>>>> that config. My vote is "no." KISS. If there is
demand for doing it at
>>>>>>> the server-group level, we'll here about it and can add
it later.
>>>>>>>
>>>>>>> As for validation, I doubt we can do much there. I think this
would be
>>>>>>> an advanced feature where users are going to have to be
careful and test
>>>>>>> things.
>>>>>>>
>>>>>>> [1] Note our schema uses lower case names with "-"
as a separator, not
>>>>>>> camel case. So "launch-command" not
"launchCommand".
>>>>>>>
>>>>>>> On 8/14/12 11:49 AM, John O'Hara wrote:
>>>>
>>>> _______________________________________________
>>>> jboss-as7-dev mailing list
>>>> jboss-as7-dev(a)lists.jboss.org
>>>> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>>>
>>> _______________________________________________
>>> jboss-as7-dev mailing list
>>> jboss-as7-dev(a)lists.jboss.org
>>> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>>>
>>
>
>
> --
> John O'Hara
> johara(a)redhat.com
>
> JBoss, by Red Hat
> Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor,
Berkshire, SI4 1TE, United Kingdom.
> Registered in UK and Wales under Company Registration No. 3798903 Directors: Michael
Cunningham (USA), Charlie Peters (USA), Matt Parsons (USA) and Brendan Lane (Ireland).
>
> _______________________________________________
> jboss-as7-dev mailing list
> jboss-as7-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
_______________________________________________
jboss-as7-dev mailing list
jboss-as7-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat
7:30 p.m.
I have updated the Javadoc in HostControllerMessages and
HostControllerLogger. HostControllerMessages is currently using id's
16500-16513, so i updated the javadoc in that class to reflect this
Latest version is here;
https://github.com/johnaoahra80/jboss-as/commit/abcaee41ba8cc4bf437ea3921...
Thanks
John
On 08/29/2012 02:39 PM, Brian Stansberry wrote:
The additional range 16500-16599 has already been allocated to these
classes, but HostControllerLogger hasn't used any in that range yet. So
all that's needed is to add javadoc and comments to
HostControllerMessages saying that class will only use 16500-16549, and
then HostControllerLogger can use 16550-16599.
On 8/29/12 4:16 AM, Kabir Khan wrote:
> Go to http://community.jboss.org/docs/DOC-16810 and reserve a new range, and be sure
to update the javadoc in both HostControllerLogger and HostControllerMessages
> On 28 Aug 2012, at 15:10, John O'Hara wrote:
--
John O'Hara
johara(a)redhat.com
JBoss, by Red Hat
Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor,
Berkshire, SI4 1TE, United Kingdom.
Registered in UK and Wales under Company Registration No. 3798903 Directors: Michael
Cunningham (USA), Charlie Peters (USA), Matt Parsons (USA) and Brendan Lane (Ireland).
4268
days inactive
4284
days old
12 comments
4 participants
participants (4)
-
Brian Stansberry -
John O'Hara -
Kabir Khan -
Misty Stanley-Jones