| Author |
Message
|
| kriersd |
 Posted: Mon Jul 14, 2003 2:10 pm Post subject: Execution Servers just die with out a trace |
 |
|
Master
Joined: 22 Jul 2002
Posts: 209
Location: IA, USA
|
I have a strange error.
Every once in a while I will notice that the Workflow server is not responding to requests. By the time I get logged into the system everything seems to be in working order again. Well, it seems that the fmcamain server has found that all the execution servers are missing and restarts them. So, I check the system over and notice that the fmcsys.log shows that the execution servers have all been started.
That's funny.. I don't see any previous error messages that would have caused all the execution servers to crash. It's really hard to debug the issue if I don't have any thing to go on.
Has anyone else seen this before, where the execution servers stop without a trace? If so how did you track down the cause?
P.S. The fmcerr.log file is empty, also!!!
Any help would be great. 
_________________
Dave Krier
IBM WebSphere MQ Workflow V3.4 Solution Designer |
|
| Back to top |
|
 |
| Ratan |
| Posted: Mon Jul 14, 2003 2:14 pm Post subject: |
|
|
Grand Master
Joined: 18 Jul 2002
Posts: 1245
|
Dave, Did you chek the system log using fmcautil?if not, try it.
Also check if there are any MQ error logs.
_________________
-Ratan |
|
| Back to top |
|
|
| kriersd |
| Posted: Wed Jul 16, 2003 8:36 am Post subject: |
|
|
Master
Joined: 22 Jul 2002
Posts: 209
Location: IA, USA
|
I did look in the fmcsys.log file, however, I did not use the fmcautil to view the log file. Doesn't the fmcautil utility just display the fmcsys.log file anyway? I guess I am not sure what if any more information would be displayed using the fmcautil utility. Can you tell me if I would see more information than what's in the fmcsys.log.
Kriersd
_________________
Dave Krier
IBM WebSphere MQ Workflow V3.4 Solution Designer |
|
| Back to top |
|
|
| Ratan |
| Posted: Wed Jul 16, 2003 8:46 am Post subject: |
|
|
Grand Master
Joined: 18 Jul 2002
Posts: 1245
|
kriersd, I am not sure if what you are saying is true. I have noticed differences in them in the past.
| Quote: |
| . Doesn't the fmcautil utility just display the fmcsys.log file anyway? |
Also you can delete your fmcsys.log file and it does not affect the log we can see using fmcautil and viceversa.
_________________
-Ratan |
|
| Back to top |
|
|
| clindsey |
| Posted: Wed Jul 16, 2003 10:04 am Post subject: |
|
|
Knight
Joined: 12 Jul 2002
Posts: 586
Location: Dallas, Tx
|
Dave,
Your execution servers are probably abending. Since they make MQ connections the MQ signal handler traps the signal and prevents the execution server from dumping core. You can be sure of this by looking in /var/mqm/errors and you should see a number of *FDC files. Grep these files for 'fmcemain' and you will problem find some results.
You can stop the workflow queue manager and restart with MQS_ACTION_ON_EXCEPTION exported in the environment before starting the queue manager. You can set this to one of two values, HANG_ALL or ABORT_ALL. If this is in production, you may want to use ABORT_ALL. This will let fmcemain dump core when it abends and then you can run pstack or dbx against the core file to get some idea what it was doing to cause the exception.
If it is in test, then use HANG_ALL. When workflow hangs, you can find out from the users what they were doing. You can also attach to the hung process with dbx to again get the current stack. You may need to open a PMR and get some help from level 2 to diagnose the output.
Hope this helps,
Charlie |
|
| Back to top |
|
|
| John Elgie |
| Posted: Tue Jul 29, 2003 2:35 pm Post subject: Execution servers just die without a trace |
|
|
Newbie
Joined: 11 Oct 2002
Posts: 3
|
I have seen errors not listed in the fmcsys.log that do appear using fmcautil. this is because fmcautil lists the rows in the system_log table rather than the flat file.
John Elgie |
|
| Back to top |
|
|
|
|
