| Author |
Message
|
| Esa |
 Posted: Wed Nov 09, 2016 2:23 am Post subject: Very slow multi-instance broker |
 |
|
Grand Master
Joined: 22 May 2008
Posts: 1387
Location: Finland
|
I have a new multi-instance broker that is very slow with deployments and mqsi commands. I think there must be something wrong with configuration. NFS4 should not make the broker that slow, I think.
In fact the broker crashes when I try to deploy a bar file. The abend file does not get created under the shared workpath but in /var/mqsi/common/errors.
Version 9.0.0.5.
mqsireportbroker shows
Work path = /var/mqsi
Shared Work Path = /mqsi/SOMEBROKER/mqsi
Work path is set by mqsiprofile script in product installation directory.
I'm just wondering if having both Work path and shared work path configured might cause slowness and crashing? |
|
| Back to top |
|
 |
| fjb_saper |
| Posted: Wed Nov 09, 2016 5:37 am Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
Thanks for specifying the version (9.0.0.5)...
As to the slowness it should not be a result of going multi-instance.
I would ask if you are deploying a SOAP node with a lot of unimplemented operations, or just a lot of operations... but then you are stating that all mqsi commands are slow...
Best bet is to take a service trace and submit it to IBM in a PMR.
Have fun 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| Craig B |
| Posted: Thu Nov 10, 2016 3:33 am Post subject: |
|
|
Partisan
Joined: 18 Jun 2003
Posts: 316
Location: UK
|
Do you know if your BAR file deploys to a none HA broker ok? Or is this the first time you have deployed this BAR file and it just happened to be a HA broker?
_________________
Regards
Craig |
|
| Back to top |
|
|
| Esa |
| Posted: Mon Nov 14, 2016 1:45 am Post subject: |
|
|
Grand Master
Joined: 22 May 2008
Posts: 1387
Location: Finland
|
I investigated the problem a little closer and found out that on both systems /var/mqm is a link to the shared nfs4 disk. That makes for example mqs.ini shared between the two queue manager instances. As far as I understand that makes the two instances the *same* instance and could be the root cause of the latency.
What do you think, should I tear down and rebuild everything? I think I can reconstruct a somewhat intact /var/mqm on both servers so that I don't need to reinstall the software. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Mon Nov 14, 2016 5:22 am Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| Esa wrote: |
I investigated the problem a little closer and found out that on both systems /var/mqm is a link to the shared nfs4 disk. That makes for example mqs.ini shared between the two queue manager instances. As far as I understand that makes the two instances the *same* instance and could be the root cause of the latency.
What do you think, should I tear down and rebuild everything? I think I can reconstruct a somewhat intact /var/mqm on both servers so that I don't need to reinstall the software. |
Try and let us know the result. How about /var/mqsi ? is that shared as well?
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| Esa |
| Posted: Mon Nov 21, 2016 12:01 pm Post subject: |
|
|
Grand Master
Joined: 22 May 2008
Posts: 1387
Location: Finland
|
| No, it did not help. And yes, /var/mqsi was shared as well, but not any more. There were some other things that I needed to fix as well but deployments remain very slow. It must be some misconfiguration of NFS4 that causes the latency. I know that our NetApp is not at the latest level, for example. |
|
| Back to top |
|
|
| mqjeff |
| Posted: Mon Nov 21, 2016 12:14 pm Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
The shared disk has to be attached to both systems at all times.
_________________
chmod -R ugo-wx / |
|
| Back to top |
|
|
| Esa |
| Posted: Wed Nov 23, 2016 9:45 am Post subject: |
|
|
Grand Master
Joined: 22 May 2008
Posts: 1387
Location: Finland
|
| The shared NFS4 disk was mounted with a bad options. |
|
| Back to top |
|
|
|
|
