pega prpc fatal error - element type a:displayname
Hi,
I am on pega prpc 7.2. We have had a few instances recently that our application had to be restarted. The one item i see in the nodeagent systemerr log just prior to it hanging is below. I was wondering if anyone had recognized anything similar, or has an idea. It looks like some tag is not closed properly, but i'm not sure why that would cause pega to hang.
Thanks for your help.
SystemErr R [Fatal Error] :4:17: Element type "a:displayname" must be followed by either attribute specifications, ">" or "/>".
Capgemini Engineering
BE
Hello,
I think node agent crash is a consequence of what's happening in the application.
When you restart the node, it then works fine?
We faced few issues like this with a kind of poison message in pr_sys_queue or a bad processing in some activities.
We work with lots of agents and when a node is stuck it's usually linked to a message from pr_sys_queue being handled and killing the node itself. Hanging or going to out of memory.
If it's detected quickly, I park this message, restart the node and it's fine.
If it takes too long to be corrected then it goes in cascading mode and node agent sometimes crash.
Regards
Anthony
US Treasury
US
thanks Anthony,
thanks for the reply. I'm assuming you're using WAS?
yes, if I restart the app its good but the users aren't happy.
I just noticed this (although the jvm memory is fine, so it seems like a temporary memory issue, but it still hangs the app...so needs a restart).
2020-09-21 10:57:15,857 [bf39b.IO.thread-in-2] [ ] [ ] ( nio.tcp.InSelectorImpl) WARN - java.lang.OutOfMemoryError: Java heap space java.lang.OutOfMemoryError : Java heap space at sun.misc…. . blockedOn at java.nio.channels.spi.AbstractInterruptibleChannel
Capgemini Engineering
BE
Hello,
so it seems my initial assumption is correct.
with OOM issue, all your users on this node are anyway stuck and cannot work and you can only restart to fix.
So you now need to find out the issue leading to OOM.
It can be a too small heap (would be the easiest one to fix) or a issue with the code/data.
Do you use agent and then could be some bad message in pr_sys_queue as already explained?
Or it's linked to user activity. They might submit data which are incorrect.
Can be too much load, bad load balancing.
Can be lots of things and it depends on your application actually.
Without details it's difficult to be more precise.
With OOM crash there's usually heap dump which can also be analysed.
Regards
Anthony
US Treasury
US
thanks Anthony
you've been very helpful.
I have a feeling its user data related, as this only started happening the last few weeks, and we are getting near fiscal year-end, so they are busy.
I will take a look at some of the items you mentioned and see if something turns up.
appreciate your help.
-Doug