[Pljava-dev] How to nicely handle Java Heap Space errors without "crashing" the PG cluster ?

Hi all, Hi Thomas,

I understand there's limited time available for this project's maintainers (or sole maintainer ?). So if you can direct me properly (if there's a possible solution to implement), I'd gladly contribute. I'd just need pointers.

When a Java VM invoked by a PL/Java call invokes a function that needs more memory than the -Xmx parameter sets, it actually crashes the PG process, which forces the whole cluster to restart.

As far as you know, is there any implementable code that would gracefully return an error to the calling PG process instead of just crashing it ?

If I understand things correctly, it would mean that the PG process survives. And as it might have to handle other PL/Java "stuff", I suppose it would have to "clean up everything" before being able to start it up all again. True ? Possible ? Other method ?

Thanks for your help/possibly implementable solution (which I would be glad to contribute if I succeed).

Fr?d?ric.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.pgfoundry.org/pipermail/pljava-dev/attachments/20141219/ec6514d1/attachment.html>