| Author |
Message
|
| chris boehnke |
 Posted: Sun Feb 06, 2011 6:49 pm Post subject: Multi instance MQ linear logging maintenance script |
 |
|
Partisan
Joined: 25 Jul 2006
Posts: 369
|
Hi,
We are using MQ 7.0.1.3 on AIX with broker 7.0.0.1 with multi instance feature for HA.
As we are handling persistent msgs, going with linear logging.
I am planning to use linear logging maintenace script, MS0L. This I found in IBM supportpacs. The 'Readme' says there is no support and we need to use on our own risk.
I am wondering whether anybody used and found any issues?.
Please let me know if anybody recommend any other linear logging maintenance script or found any issues with MS0L?.
Thanks, |
|
| Back to top |
|
 |
| fjb_saper |
| Posted: Sun Feb 06, 2011 8:03 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
I'm using MS0L very happily on unix.
a few precautions are needed
- make sure to use the temporary file option (copy of *.LOG files of the error directory. This avoids the trouble happening if the files are switched while you're looking for the lowest needed log file)
- Usually I symlink the logs to /var/mqm/log
- make sure you use the latest version. Some of the earlier version (V5.3) do not handle the fact of the additional space (or missing space) in the date timestamps template (v6)...
I'm sure if you do find a specific problem and bring it to the attention of the author you may/will get some attention.
Have fun 
_________________
MQ & Broker admin
Last edited by fjb_saper on Mon Feb 07, 2011 2:39 pm; edited 1 time in total |
|
| Back to top |
|
|
| Mirage |
| Posted: Mon Feb 07, 2011 1:23 pm Post subject: |
|
|
Apprentice
Joined: 30 Nov 2004
Posts: 44
Location: IL
|
We are also using this service pack in our environment (AIX) for the last 3/5 years and are happy with it.
To your question, we recently created a Multi-Instance queue manager in our DEV environment and in order to use this script, we had to create Two symbolic links. The script was failing since it was looking for the log and error files under the default /var/mqm/ prefix.
Error:
cleanmqlogs: error opening directory /var/mqm/log/<QMGR>/active: A file or directory in the path name does not exist.
Symbolic Links:
1. /var/mqm/qmgrs/<QMGR>/errors --> <contact admin Share>/<QMGR>/errors
2. /var/mqm/log/<QMGR>/active --> <contact admin Share>/<QMGR>/active
So far the script is running fine after adding the links.
Cheers,
Raghu |
|
| Back to top |
|
|
| chris boehnke |
| Posted: Mon Feb 07, 2011 5:17 pm Post subject: |
|
|
Partisan
Joined: 25 Jul 2006
Posts: 369
|
Thanks for sharing the info.
Are you guys using archive or purge option?. Does this script ever needed archived logs to recover as you guys are using since long time?.
Which option is recommended. If Archive, how old are you saving/ keeping the archived logs?.
Thanks. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Mon Feb 07, 2011 10:31 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| chris boehnke wrote: |
Thanks for sharing the info.
Are you guys using archive or purge option?. Does this script ever needed archived logs to recover as you guys are using since long time?.
Which option is recommended. If Archive, how old are you saving/ keeping the archived logs?.
Thanks. |
I archive and purge the archive after x days. x will depend on your needs and disk space.
So far I have had 1 incident, in the last 3 years, where for whatever reason the logs had been destroyed / missing (I could not eliminate human intervention as a cause..., like some overzealous OS admin deleting logs to provide more space...). The hack of recreating the logs with the same characteristics and copying them worked fine in that case. Of course you restart with the sequence numbers...
Just make sure you have enough log space for 1.5 days. In some cases I had to bounce a qmgr to free logs (prob. with uncommitted transactions => long running trans) so that they can be archived. You also need to have OS alerts on disk space over 50% and 75% full. If the batch rcdmqimg failed to free old restart logs, you may need to bounce the corresponding qmgr.
We record the image once a day through a crontab triggered script.
We run the archive process once an hour. Most of the time we just get an entry in the log saying nothing to archive and after the run of rcdmqimg we see the logs getting archived when the archive program runs...
This allows for archiving if the qmgr frees the logs before the record image is being run...
In normal operations I don't expect you to have to restore any logs from archive ever, as the needed logs for restart and media recovery are not archived from the active directory....
Have fun
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| chris boehnke |
| Posted: Tue Feb 08, 2011 6:10 am Post subject: |
|
|
Partisan
Joined: 25 Jul 2006
Posts: 369
|
| Mirage wrote: |
We are also using this service pack in our environment (AIX) for the last 3/5 years and are happy with it.
To your question, we recently created a Multi-Instance queue manager in our DEV environment and in order to use this script, we had to create Two symbolic links. The script was failing since it was looking for the log and error files under the default /var/mqm/ prefix.
Error:
cleanmqlogs: error opening directory /var/mqm/log/<QMGR>/active: A file or directory in the path name does not exist.
Symbolic Links:
1. /var/mqm/qmgrs/<QMGR>/errors --> <contact admin Share>/<QMGR>/errors
2. /var/mqm/log/<QMGR>/active --> <contact admin Share>/<QMGR>/active
So far the script is running fine after adding the links.
Cheers,
Raghu |
Thanks for sharing the info.
I modified mqs.ini instead of establishing softlinks. Here is what I am doing and it works for me.
| Quote: |
#********************************************************************#
#* *#
#* <N_OCO_COPYRIGHT> *#
#* Licensed Materials - Property of IBM *#
#* *#
#* 63H9336 *#
#* (C) Copyright IBM Corporation 1994, 2003 *#
#* *#
#* <NOC_COPYRIGHT> *#
#* *#
#********************************************************************#
#***********************************************************************#
#* Module Name: mqs.ini *#
#* Type : WebSphere MQ Machine-wide Configuration File *#
#* Function : Define WebSphere MQ resources for an entire machine *#
#* *#
#***********************************************************************#
#* Notes : *#
#* 1) This is the installation time default configuration *#
#* *#
#***********************************************************************#
AllQueueManagers:
#********************************************************************#
#* The path to the qmgrs directory, below which queue manager data *#
#* is stored *#
#********************************************************************#
DefaultPrefix=<NFS mount>/wmq/data
LogDefaults:
LogDefaultPath=/var/mqm/log
QueueManager:
Name=QM1
Prefix=<NFS mount>/wmq/data
Directory=QM1
DataPath=<NFS mount>/wmq/data/qmgrs/QM1
QueueManager:
Name=QM2
Prefix=/var/mqm
Directory=QM2
DataPath=<NFS mount>/wmq/data/qmgrs/QM2 |
I set the 'DefaultPrefix=<NFS mount>/wmq/data' in mqs.ini. I hope this will eliminates the softlinks from /var/mqm/log, /var/mqm/errors to contact admin.
Let me know your thoughts.
Thanks |
|
| Back to top |
|
|
| Mirage |
| Posted: Tue Feb 08, 2011 7:23 am Post subject: |
|
|
Apprentice
Joined: 30 Nov 2004
Posts: 44
Location: IL
|
I also thought of changing the default prefix in mqs.ini file before going with the symbolic links. But, when I verified the cleanmqlogs perl script, I could see "qmgrs" folder name hard coded for the path within the script. Since we did not include "qmgrs" in our contact admin share folder structure, I didn't try updating queue manager prefix.
I do see your contact admin share has "qmgrs" in the path. Maybe thats the reason why its working for you.
I'll try in the Lab environment without "qmgrs" in the path and let you know how it goes.
Thanks. |
|
| Back to top |
|
|
| chris boehnke |
| Posted: Tue Feb 08, 2011 8:55 am Post subject: |
|
|
Partisan
Joined: 25 Jul 2006
Posts: 369
|
| Mirage wrote: |
I also thought of changing the default prefix in mqs.ini file before going with the symbolic links. But, when I verified the cleanmqlogs perl script, I could see "qmgrs" folder name hard coded for the path within the script. Since we did not include "qmgrs" in our contact admin share folder structure, I didn't try updating queue manager prefix.
I do see your contact admin share has "qmgrs" in the path. Maybe thats the reason why its working for you.
I'll try in the Lab environment without "qmgrs" in the path and let you know how it goes.
Thanks. |
Yes. You are right. The script is hardcoded with <contact admin>/qmgrs.
Thatswhy I created the qmgrs folder under contact admin storage and then created the Qmgrs to use shared storage.
The script MS0L is expecting QMgrs directory and we need to have that. If the softlink works for you, its cool..no need to change mqs.ini.
Let me know if anybody finds anything wrong with my approach of editing mqs.ini file.
Thanks. |
|
| Back to top |
|
|
| Mirage |
| Posted: Tue Feb 08, 2011 11:30 am Post subject: |
|
|
Apprentice
Joined: 30 Nov 2004
Posts: 44
Location: IL
|
Hmm...I just checked the support pack MS62 on IBM site and the version we have in our environment is outdated. 
Ours is V 1.19 (from year 1999  can't believe its still working.)
We may have to upgrade to the latest version (which is reading both mqs.ini and qm.ini for the data and log path). The old version has both these paths hardcoded and hence the need for our symbolic links.
Thanks. |
|
| Back to top |
|
|
|
|
