| Author |
Message
|
| mq_pm |
 Posted: Thu Jun 24, 2004 11:02 am Post subject: is it safe to kill broker processes??? |
 |
|
Centurion
Joined: 27 Aug 2003
Posts: 132
|
Hi,
I am trying to recycle broker and it is not shutting down. Is it safe to kill using kill -9 , the processes that were not shutting down and then do normal recycle. Will it cause any issues doing this way???
Any suggestion is appreciated.
Thanks. |
|
| Back to top |
|
 |
| kirani |
| Posted: Thu Jun 24, 2004 11:07 am Post subject: |
|
|
Jedi Knight
Joined: 05 Sep 2001
Posts: 3779
Location: Torrance, CA, USA
|
You can try to find out which EG is not allowing the broker to shutdown and try killing that EG process only. If you cannot find it then you can use [b]mqsistop -i[b] command to shutdown the broker. Be very careful when doing this in Prod environment.
_________________
Kiran
IBM Cert. Solution Designer & System Administrator - WBIMB V5
IBM Cert. Solutions Expert - WMQI
IBM Cert. Specialist - WMQI, MQSeries
IBM Cert. Developer - MQSeries
|
|
| Back to top |
|
|
| PGoodhart |
| Posted: Thu Jun 24, 2004 11:09 am Post subject: |
|
|
Master
Joined: 17 Jun 2004
Posts: 278
Location: Harrisburg PA
|
Have you tried to force the mqsistop using the -i flag?
If that doesn't work then I have used the kill -9 command before and it doesn't hurt anything. You may have to stop the broker after you kill the processes again, and you will get entries in the error log (not that it matters).
_________________
Patrick Goodhart
MQ Admin/Web Developer/Consultant
WebSphere Application Server Admin |
|
| Back to top |
|
|
| mq_pm |
| Posted: Thu Jun 24, 2004 11:23 am Post subject: |
|
|
Centurion
Joined: 27 Aug 2003
Posts: 132
|
Thanks for responding. This is in Production only. Yes we tried using -i option but still problem exists. So killing the broker process to stop the broker and then kill all the hanging semaphores etc. So it won't hurt doing this????
Thanks. |
|
| Back to top |
|
|
| PGoodhart |
| Posted: Thu Jun 24, 2004 11:28 am Post subject: |
|
|
Master
Joined: 17 Jun 2004
Posts: 278
Location: Harrisburg PA
|
Not a thing really... You may want to try and stop the Queue manager first.
_________________
Patrick Goodhart
MQ Admin/Web Developer/Consultant
WebSphere Application Server Admin |
|
| Back to top |
|
|
| MikeC |
| Posted: Fri Jun 25, 2004 7:35 am Post subject: |
|
|
Acolyte
Joined: 30 Jun 2003
Posts: 55
Location: Toronto, Canada
|
Be very careful killing semaphores and shared memory segments. We had a line in our shutdown script to do exactly that. Everything seemed fine at first. But after a while (several stops and starts) it began to corrupt something in the broker environment. Soon all our messages got dumped in the failure queue. It took ages to determine why it was happening.  But after removing the killing of semaphores and shared memory, the problem went away and never came back.
To me, killing shared memory segments and semaphores is an absolute last resort. .... I'd focus my attention on why the broker isn't coming down cleanly using normal mqsistop. .. Or at the most mqsistop -i
Cheers.
_________________
-Mike. |
|
| Back to top |
|
|
| Heinz57 |
| Posted: Mon Jun 28, 2004 7:59 am Post subject: |
|
|
Apprentice
Joined: 21 Jan 2004
Posts: 26
Location: Syracuse, NY
|
| Are you trying to stop and start the broker or just refresh the DFE's ? If you're just trying to refresh, you can use the mqsireload <broker>. We use this in production daily to stop and start the data flow engines and refresh the odbc connections. We do have a fail over though since this refresh would cause a production outage. |
|
| Back to top |
|
|
| tdolby |
| Posted: Tue Jun 29, 2004 1:07 am Post subject: |
|
|
Newbie
Joined: 29 Jun 2004
Posts: 8
|
In general, killing broker processes is not nice but no worse than mqsistop -i, which sends SIGKILL (-9) under the covers. I think there is some confusion surrounding semaphores and shared memory, though - the broker only uses a couple of semaphores, and they are automatically reset (though not deleted) by the O.S. when a process dies (using SEM_UNDO). The other semaphores and shared memory segments are probably MQ ones - deleting them without ensuring all MQ processes have terminated could lead to sub-optimal results 
The following should be reasonably safe (though step 3 is probably unsupported):
1) Do an mqsistop, and see if things shut down normally
2) If they haven't, do an mqsistop -i
3) If any DFE or bip* processes survive the stop -i, shoot them
4) If possible, end the MQ queue manager, and restart it.
There are also a few more debugging steps we could try, but they could take some effort.
Also, there's no platform mentioned in the original post - where is this happening?
Surgeon General's Warning: Killing processes may cause data and/or hair loss. Your mileage may vary. Not intended for use by children under the age of three. Your home is at risk if you do not keep up payments on a mortgage or other loan secured on it. If the product comes into contact with skin or eyes, flush with water and call a doctor immediately
_________________
Trevor Dolby
IIB Development |
|
| Back to top |
|
|
| chanduy9 |
| Posted: Tue Jun 29, 2004 7:47 am Post subject: |
|
|
Disciple
Joined: 28 Nov 2001
Posts: 177
Location: USA
|
Hi,
See if there is any EG are spinning...if any.. kill the EG..you can do this by seeing which DataFlowEngine taking more CPU time.
Where is your broker running...??
Thanks,
Chandra. |
|
| Back to top |
|
|
|
|
