You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@uima.apache.org by "Jörn Kottmann (JIRA)" <de...@uima.apache.org> on 2010/05/18 18:34:44 UTC
[jira] Created: (UIMA-1789) deployAsyncService script should use
-XX:+HeapDumpOnOutOfMemoryError
deployAsyncService script should use -XX:+HeapDumpOnOutOfMemoryError
--------------------------------------------------------------------
Key: UIMA-1789
URL: https://issues.apache.org/jira/browse/UIMA-1789
Project: UIMA
Issue Type: Improvement
Components: Async Scaleout
Reporter: Jörn Kottmann
Priority: Minor
In case a user application needs to much memory an OuOfMemoryError stops the UIMA AS service. It would
be nice if the VM would just dump the memory for further analysis and stops afterwards.
For a sun vm this can be achieved with this command line option:
-XX:+HeapDumpOnOutOfMemoryError
Can this be done in our script, does it also work with other VMs ?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (UIMA-1789) deployAsyncService script should use
-XX:+HeapDumpOnOutOfMemoryError
Posted by "Eddie Epstein (JIRA)" <de...@uima.apache.org>.
[ https://issues.apache.org/jira/browse/UIMA-1789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12873996#action_12873996 ]
Eddie Epstein commented on UIMA-1789:
-------------------------------------
A problem with enabling heap dumps as the default is that many large files can accumulate for unsuspecting users. To me, enabling heaps dumps is something done specifically when looking for problems, not a typical default configuration. It would be nice to hear other opinions on the matter.
> deployAsyncService script should use -XX:+HeapDumpOnOutOfMemoryError
> --------------------------------------------------------------------
>
> Key: UIMA-1789
> URL: https://issues.apache.org/jira/browse/UIMA-1789
> Project: UIMA
> Issue Type: Improvement
> Components: Async Scaleout
> Reporter: Jörn Kottmann
> Priority: Minor
>
> In case a user application needs to much memory an OuOfMemoryError stops the UIMA AS service. It would
> be nice if the VM would just dump the memory for further analysis and stops afterwards.
> For a sun vm this can be achieved with this command line option:
> -XX:+HeapDumpOnOutOfMemoryError
> Can this be done in our script, does it also work with other VMs ?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (UIMA-1789) deployAsyncService script should use
-XX:+HeapDumpOnOutOfMemoryError
Posted by "Eddie Epstein (JIRA)" <de...@uima.apache.org>.
[ https://issues.apache.org/jira/browse/UIMA-1789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12872341#action_12872341 ]
Eddie Epstein commented on UIMA-1789:
-------------------------------------
The process (UIMA service) created by deployAsyncService picks up custom JVM parameters from the environmental parameter UIMA_JVM_OPTS. If this parameter does not exist, the default is "-Xms128M -Xmx800M".
Does this resolve the issue?
> deployAsyncService script should use -XX:+HeapDumpOnOutOfMemoryError
> --------------------------------------------------------------------
>
> Key: UIMA-1789
> URL: https://issues.apache.org/jira/browse/UIMA-1789
> Project: UIMA
> Issue Type: Improvement
> Components: Async Scaleout
> Reporter: Jörn Kottmann
> Priority: Minor
>
> In case a user application needs to much memory an OuOfMemoryError stops the UIMA AS service. It would
> be nice if the VM would just dump the memory for further analysis and stops afterwards.
> For a sun vm this can be achieved with this command line option:
> -XX:+HeapDumpOnOutOfMemoryError
> Can this be done in our script, does it also work with other VMs ?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Closed: (UIMA-1789) deployAsyncService script should use
-XX:+HeapDumpOnOutOfMemoryError
Posted by "Jörn Kottmann (JIRA)" <de...@uima.apache.org>.
[ https://issues.apache.org/jira/browse/UIMA-1789?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jörn Kottmann closed UIMA-1789.
-------------------------------
Resolution: Won't Fix
Enabling the -XX:+HeapDumpOnOutOfMemoryError option by default will not work on all VMs. Enabling this feature also might confuse users which do not expect a memory dump, especially during testing/development of an AE where OOME might occur frequently.
For everyone who wants to enable this option set the following environment variable:
UIMA_JVM_OPTS=-XX:+HeapDumpOnOutOfMemoryError
Depending on the situation it is useful to diagnose memory leaks in a production environment.
> deployAsyncService script should use -XX:+HeapDumpOnOutOfMemoryError
> --------------------------------------------------------------------
>
> Key: UIMA-1789
> URL: https://issues.apache.org/jira/browse/UIMA-1789
> Project: UIMA
> Issue Type: Improvement
> Components: Async Scaleout
> Reporter: Jörn Kottmann
> Priority: Minor
>
> In case a user application needs to much memory an OuOfMemoryError stops the UIMA AS service. It would
> be nice if the VM would just dump the memory for further analysis and stops afterwards.
> For a sun vm this can be achieved with this command line option:
> -XX:+HeapDumpOnOutOfMemoryError
> Can this be done in our script, does it also work with other VMs ?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (UIMA-1789) deployAsyncService script should use
-XX:+HeapDumpOnOutOfMemoryError
Posted by "Marshall Schor (JIRA)" <de...@uima.apache.org>.
[ https://issues.apache.org/jira/browse/UIMA-1789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12874331#action_12874331 ]
Marshall Schor commented on UIMA-1789:
--------------------------------------
Another potential issue is that this parameter is one of the XX ones, which means it is likely not standard for all JVMs. For instance, the IBM JVM doesn't have this option - see http://publib.boulder.ibm.com/infocenter/javasdk/v6r0/index.jsp and search on -XX.
Another danger for users is that they would not expect this to be enabled by default. This means that they could become puzzled by very long termination events (in the case they were using a 64-bit OS and running with very large heaps), or run out of disk space, etc.
My opinion is that the right diagnosis strategy for an OutOfMemory error is probably situation dependent, and only some of them (on some JVMs) would warrant a heap dump. So I think this should be left to developers to specify explicitly (if their JVM supports it), and not enabled by default.
> deployAsyncService script should use -XX:+HeapDumpOnOutOfMemoryError
> --------------------------------------------------------------------
>
> Key: UIMA-1789
> URL: https://issues.apache.org/jira/browse/UIMA-1789
> Project: UIMA
> Issue Type: Improvement
> Components: Async Scaleout
> Reporter: Jörn Kottmann
> Priority: Minor
>
> In case a user application needs to much memory an OuOfMemoryError stops the UIMA AS service. It would
> be nice if the VM would just dump the memory for further analysis and stops afterwards.
> For a sun vm this can be achieved with this command line option:
> -XX:+HeapDumpOnOutOfMemoryError
> Can this be done in our script, does it also work with other VMs ?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (UIMA-1789) deployAsyncService script should use
-XX:+HeapDumpOnOutOfMemoryError
Posted by "Jörn Kottmann (JIRA)" <de...@uima.apache.org>.
[ https://issues.apache.org/jira/browse/UIMA-1789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12873594#action_12873594 ]
Jörn Kottmann commented on UIMA-1789:
-------------------------------------
Using UIMA_JVM_OPTS to set -XX:+HeapDumpOnOutOfMemoryError would
still require the user to set the environment variable. I would like to have this
option set as a default for everyone.
What do you think about that ? Is that possible or does it have any negative side effects ?
> deployAsyncService script should use -XX:+HeapDumpOnOutOfMemoryError
> --------------------------------------------------------------------
>
> Key: UIMA-1789
> URL: https://issues.apache.org/jira/browse/UIMA-1789
> Project: UIMA
> Issue Type: Improvement
> Components: Async Scaleout
> Reporter: Jörn Kottmann
> Priority: Minor
>
> In case a user application needs to much memory an OuOfMemoryError stops the UIMA AS service. It would
> be nice if the VM would just dump the memory for further analysis and stops afterwards.
> For a sun vm this can be achieved with this command line option:
> -XX:+HeapDumpOnOutOfMemoryError
> Can this be done in our script, does it also work with other VMs ?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (UIMA-1789) deployAsyncService script should use
-XX:+HeapDumpOnOutOfMemoryError
Posted by "Eddie Epstein (JIRA)" <de...@uima.apache.org>.
[ https://issues.apache.org/jira/browse/UIMA-1789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12873287#action_12873287 ]
Eddie Epstein commented on UIMA-1789:
-------------------------------------
Sorry, I am confused. Before running the script one can set
UIMA_JVM_OPTS=-XX:+HeapDumpOnOutOfMemoryError
Why isn't this the option requested?
> deployAsyncService script should use -XX:+HeapDumpOnOutOfMemoryError
> --------------------------------------------------------------------
>
> Key: UIMA-1789
> URL: https://issues.apache.org/jira/browse/UIMA-1789
> Project: UIMA
> Issue Type: Improvement
> Components: Async Scaleout
> Reporter: Jörn Kottmann
> Priority: Minor
>
> In case a user application needs to much memory an OuOfMemoryError stops the UIMA AS service. It would
> be nice if the VM would just dump the memory for further analysis and stops afterwards.
> For a sun vm this can be achieved with this command line option:
> -XX:+HeapDumpOnOutOfMemoryError
> Can this be done in our script, does it also work with other VMs ?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (UIMA-1789) deployAsyncService script should use
-XX:+HeapDumpOnOutOfMemoryError
Posted by "Jörn Kottmann (JIRA)" <de...@uima.apache.org>.
[ https://issues.apache.org/jira/browse/UIMA-1789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12872672#action_12872672 ]
Jörn Kottmann commented on UIMA-1789:
-------------------------------------
No, dumping the memory when an OutOfMemoryError occurs makes it possible to analyze the issue
afterwards, if the system is set up correctly (so its not crashing directly after startup) a memory leak can cause an OOME after maybe a few weeks of operation. In my opinion thats why you want to dump the memory on an OOME in a production environment, when the system fails its possible
to analyze the cause directly afterwards. If its not dumped, you maybe have to wait for a few days until its clear in the memory dump (e.g. taken with jmap or gcore), where the leak is.
If its possible it would be nice to have this option in the script as a default.
> deployAsyncService script should use -XX:+HeapDumpOnOutOfMemoryError
> --------------------------------------------------------------------
>
> Key: UIMA-1789
> URL: https://issues.apache.org/jira/browse/UIMA-1789
> Project: UIMA
> Issue Type: Improvement
> Components: Async Scaleout
> Reporter: Jörn Kottmann
> Priority: Minor
>
> In case a user application needs to much memory an OuOfMemoryError stops the UIMA AS service. It would
> be nice if the VM would just dump the memory for further analysis and stops afterwards.
> For a sun vm this can be achieved with this command line option:
> -XX:+HeapDumpOnOutOfMemoryError
> Can this be done in our script, does it also work with other VMs ?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.