| Author |
Message
|
| alphamale |
 Posted: Fri Mar 10, 2006 8:22 am Post subject: Monitoring of MQ on Linux using Veritas Cluster |
 |
|
Novice
Joined: 09 Feb 2005
Posts: 12
Location: UK
|
Anyone out there know a clever way to monitor MQ on a RedHat Linux box using a Veritas Cluster ?
I was going to check for the existence of key processes (e.g. amqzllp0), but on my installation there are multiple instances for each of them (something to do with how Linux does threads ?).
Since I'm not sure if the number of instances of a process is static forever, I don't simply want to just count how many I find.
I also don't want to assume that if one exists, then all is well, just in case one instance could die and leave the rest running (but damage the QM).
I've seen someone just use the 'dspqm' command and check for 'running', but I'm sure that's not foolproof and doesn't cover things like missing listeners etc.
If anyone has come up with a clever, reliable solution I'd love to hear it 
Below is a 'ps -ef' so you can see what I get for one MQ process
mqm 6921 1 0 Feb27 ? 00:00:09 amqzxma0 -m GBSCC01D
mqm 6922 6921 0 Feb27 ? 00:00:09 amqzxma0 -m GBSCC01D
mqm 6924 6922 0 Feb27 ? 00:00:06 amqzxma0 -m GBSCC01D
mqm 6926 6922 0 Feb27 ? 00:00:00 amqzxma0 -m GBSCC01D
mqm 6927 6922 0 Feb27 ? 00:00:03 amqzxma0 -m GBSCC01D
mqm 6950 6922 0 Feb27 ? 00:00:01 amqzxma0 -m GBSCC01D
mqm 6951 6922 0 Feb27 ? 00:00:00 amqzxma0 -m GBSCC01D
I've done this before on Solaris where there is just one instance of each process and it detected errors perfectly, but Linux has me stumped. |
|
| Back to top |
|
 |
| jefflowrey |
| Posted: Fri Mar 10, 2006 8:26 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
As long as you find at least one instance of a process, do you really care how many there are?
Also, I would review the support packs on MQ and Veritas. I think the only one is for Solaris, but most of it should be applicable.
Also, what you are seeing is not multiple processes, but rather Linux is reporting each thread in an app as a process.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| alphamale |
| Posted: Fri Mar 10, 2006 8:39 am Post subject: |
|
|
Novice
Joined: 09 Feb 2005
Posts: 12
Location: UK
|
I'd be happy to test for just one instance if I could be sure that it meant the MQ component was working.
However, what happens if one thread dies - do the others die too on Linux, or could it leave its siblings behind and leave the MQ component is an error-state ?
I know I'm probably asking for too much, but I need to find as foolproof a solution as I can.
Have checked the support packs, but can't see anything clarifying my issue (and this reporting of threads as processes seems to be a Linux phenomenon). |
|
| Back to top |
|
|
| wschutz |
| Posted: Fri Mar 10, 2006 9:54 am Post subject: |
|
|
Jedi Knight
Joined: 02 Jun 2005
Posts: 3316
Location: IBM (retired)
|
you could hack a script that used amqsputC to put to SYSTEM.DEFAULT.LOCAL.QUEUE and see if you get that message off.... if that works, you'd get a certain level of comfort.....
_________________
-wayne |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Fri Mar 10, 2006 9:59 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
It seems to me that amqsCNXC would be more useful for that...
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| markt |
| Posted: Fri Mar 10, 2006 10:08 am Post subject: |
|
|
Knight
Joined: 14 May 2002
Posts: 508
|
The non-existence of processes will tell you that a qmgr is not running.
But the existence of those processes tells you nothing definitive about the health of a qmgr.
The only way to tell if a qmgr is running is to do some work to it - the HA supportpac scripts use PING QMGR through runmqsc. |
|
| Back to top |
|
|
|
|
