Register or Login to Your Account
I would like to share my experiences in my project where we are using webMethods 8.x IS with SP1 and engineering fix, we were also using MWS and optimise.
Recently we had heavy data load from source applications and also we migrated bunch of new application to webMethods IS, before migration we had everything working fine except for a few ftp exceptions due to target server unavailabilitys which were retried or resubmitted. The recent migration had lot of issues on boaard, some of them were
1) IS crashing with Java errors and memory exceptions. We tried increasing MAX_Perm size parameter but didn't help.
[b]There is insufficient memory for the Java Runtime Environment to continue. Native memory allocation (malloc) failed to allocate 92016 bytes for Arena::Amalloc[/b]
From my point of view and based on Java logs, the error occurs because the JIT compiler could not allocate enough C-Heap. We tried to reduce the Java-Heap and Perm-Gen size, so more C-Heap is available and we might be lucky that this is enough.
In order to tune our jvm we have done following settings:[/b]-[b] Most probably our configured swap space was too small. So, we set the parameters -XX:+UseDefaultStackSize -Xss256K[/b]
- [b]Set -Xmn (the size of the heap for the young generation) value to a lower than the -Xmx value[/b]
-[b] Reduce Perm-Gem size (-XX:MaxPermSize=1G) to a lower value[/b]
- We also have to move from 32 bit to 64-bit JVM
- Since this is pretty much related to the JVM We have considered to update our JVM to the latest available build.
But still the Issue was not resolved.
2) We also had some scheduler issues error message appear in the IS server log and error log [[b]wm-cjdbc40-0034][Oracle JDBC Driver]Row update failed.[/b] not updating the next run interval because of the exception returned by the scheduler and the exception text/length was increasing the max permisible size at the scheduler DB, so we increased the column length to store large error texts as workaround. But still we had problems with row update and SAG suggested [b]DataDirect fix - SCG_8.0.1_DataDirect4.0_Fix1 [/b]
This happened suddenly and restarting db server should resolve the issue. The reason is the table from database is being used so the IS could not update the table. When db restarted, db is in initial state, so the table could be updated. Since huge no of connections to DB could also cause the issue.
After DataDirect fix installation the scheduler issues seems to be gone, but after a fortnight we were again facing scheduler issues. Still the problems with schedulers were not solved and we are in touch for the solving them.
For Java exceptions, these kind of problems are more likely to occur on 32-bit systems. If the JVM heap size is set too big, there may not be enough native memory remaining for our application to run.
When an application runs out of native memory, the JVM heap may still have plenty of space available. To address this problem, we can reduce our JVM Max heap size. By doing so, we will allow more space for the native memory and this looks like normal GC operation.
In our IS server.log there's no indication of running short of memory either, the only thing of concern is repeated socket timeouts on the broker connection suggesting network problems.
The final fix SAG suggested containing PIE-19689 is IS_8.0_SP1_Core_Fix20 which was released some time ago so running with engineering fix IS_8.0_SP1_Core_PIE_19689_Engg_Fix2 is no longer supported for us.
Then we have installed core fix 22 and JVM 64 bit on our systems.
After this installation we do not have any Java related issues on board.
I will keep posted about the latest developments in our project, anyone can comment about their developments and experiences with webMethods 8.x
Thank you for sharing useful information.