ASG
IBM
Zystems
Cressida
Icon
Netflexity
 
  MQSeries.net
Search  Search       Tech Exchange      Education      Certifications      Library      Info Center      SupportPacs      LinkedIn  Search  Search                                                                   FAQ  FAQ   Usergroups  Usergroups
 
Register  ::  Log in Log in to check your private messages
 
RSS Feed - WebSphere MQ Support RSS Feed - Message Broker Support

MQSeries.net Forum Index » General IBM MQ Support » AMQ7469: Transactions rolled back to release log space.

Post new topic  Reply to topic
 AMQ7469: Transactions rolled back to release log space. « View previous topic :: View next topic » 
Author Message
gbailey
PostPosted: Wed Dec 22, 2004 9:11 am    Post subject: AMQ7469: Transactions rolled back to release log space. Reply with quote

Newbie

Joined: 25 Sep 2002
Posts: 9
Location: Farnborough, UK

Now this seems to be a frequently raised subject on MQSeries.net, however I am going to have to bring it up again.

We noticed, by looking in our AMQERR*.LOG files that this error (AMQ7469) was being written every five minutes. Unfortunately the log files had wrapped round since it started, so we have no idea how long it had been happening for.

Reasearch on the forum and in the Systems Admin guide suggested that this was due to one transaction requring more log space than was available. This was suprising in itself, as although we only had the default circular log files (3 primary, 2 secondary, 4MB), all the messages that come through any channel should be quite small (<1KB). We get quite a lot (~120 a second), and the total through the system is no more than ~400 a second, but the batch sizes are 50 on all our receiver channels. Fifty messages should be short of the maximum of 20MB our log set up allows for. Is this too simplistic an approach; can the sheer volume of messages make a difference - especially with the queue manager worrying about more than one channel at a time?

Since we had a large amount of free space we changed the number of log files to the maximum allowed (we used 32 primary, 31 secondary) and restarted the queue manager. The effect of this was that the AMQ7469 error now occurs every 63 minutes (is this number a coincidence?).

So my first question is, has any one got any idea what is going on? Do I have some massive transaction trying to commit that is never going to have enough log space? Or is this related to the volume of different messages coming through the queue manager?

Secondly, only 51 log files are being written, even though we have defined 63. Why is this?

What happens when the transaction gets rolled back? Is it placed on the original transmit queue again? If so does it then try and resend again? We can find no build up of messages anywhere on our system, and none of our connecting systems are reporting any errors.
Back to top
View user's profile Send private message
csmith28
PostPosted: Wed Dec 22, 2004 9:26 am    Post subject: Reply with quote

Grand Master

Joined: 15 Jul 2003
Posts: 1196
Location: Arizona

How big are your current logs? I am guessing 4meg.

Platform: AIX5.1 WMQ5.3.0.6
I encountered the 51 logs being written as well. Someone had edited the qm.ini to 60 primary and 3 secondary but only 51 primarys were created. IBM seemed to be aware of this but since I was planning to recreate the MQManger in question increasing the logfile size from 4meg to 64meg we did not get into that.

First thing I would do is try editing your qm.ini file again:

LogPrimaryFiles = 51
LogSecondaryFiles = 12

If that doesn't help you will want to consider re-creating your MQManager to increase the size of the logs or reduce the amount of persistent messages the MQManager is handling.

The WMQ Administration Guide gives a good description of what is happening when transactions get rolled back. Yes the MQManager will attempt to resend persistent messages.
_________________
Yes, I am an agent of Satan but my duties are largely ceremonial.
Back to top
View user's profile Send private message
csmith28
PostPosted: Wed Dec 22, 2004 9:40 am    Post subject: Reply with quote

Grand Master

Joined: 15 Jul 2003
Posts: 1196
Location: Arizona

IBM Also suggested I do the following. Note this is only the case for AIX Servers that are Dedicated to being an MQSeries Server, or so I was told:
Quote:

Tuning performance for nonpersistent messages on AIX
If you are using AIX, consider setting your tuning parameter to exploit full
performance for nonpersistent messages. To set this tuning parameter, a root user
must issue the command /usr/samples/kernal/vmtune -c 0.
The effect of this command persists until the next reboot. An AIX administrator
can add a line such as vmtune:2:once:/usr/samples/kernal/vmtune -c 0 to the
system /etc/inittab file to cause the command to be issued on every reboot. This
command is made available by installing the bos.adt.samples fileset from the AIX
installation CDs.
Normally, nonpersistent messages are kept only in memory, but there are
circumstances where AIX can schedule nonpersistent messages to be written to
disk. Messages scheduled to be written to disk are unavailable for MQGET until
the disk write completes. The suggested tuning command varies this threshold;
instead of scheduling messages to be written to disk when 16 kilobytes of data are
queued, the write-to-disk occurs only when real storage on the machine becomes
close to full.


Per above I was told that in many instances even a non-persistent message will be written to the active logs and changing the vmtune paramitter will prevent it from happening unless the server starts to run out of memory.
_________________
Yes, I am an agent of Satan but my duties are largely ceremonial.
Back to top
View user's profile Send private message
Nigelg
PostPosted: Wed Dec 22, 2004 11:27 pm    Post subject: Reply with quote

Grand Master

Joined: 02 Aug 2004
Posts: 1046

The msg does not mean that one trasnsaction needs more than the entire log space, just more than the remaining log space. Log space is not recovered until a checkpoint; this is done every 10000 queue operations, so there may not be enough headroom in the log for the transaction. A checkpoint is also done when a log is getting full, but it may not complete in time to prevent the msg to appear.
The obvious recommendation is to increase the size of the log. If a qmgr is created with the largest possible log file size, 16384 blocks, it is easier to increase the log space.
Non-persistent msgs are never written to the log. Th disk write referred to is to the queue file.
Back to top
View user's profile Send private message
gbailey
PostPosted: Thu Dec 23, 2004 2:55 am    Post subject: Reply with quote

Newbie

Joined: 25 Sep 2002
Posts: 9
Location: Farnborough, UK

Thank you for the advice and the explanation. For the record, we are running on AIX5.2 with WMQ5.3.0.6. The machines also run WBIMB.

We are going to try changing the qm.ini as suggested, and if that doesn't work resort to rebuilding the queue managers.
Back to top
View user's profile Send private message
Display posts from previous:   
Post new topic  Reply to topic Page 1 of 1

MQSeries.net Forum Index » General IBM MQ Support » AMQ7469: Transactions rolled back to release log space.
Jump to:  



You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
Protected by Anti-Spam ACP
 
 


Theme by Dustin Baccetti
Powered by phpBB © 2001, 2002 phpBB Group

Copyright © MQSeries.net. All rights reserved.