| Author |
Message
|
| KeeferG |
 Posted: Tue Nov 16, 2004 8:01 am Post subject: amqzlaa0 process and fip |
 |
|
Master
Joined: 15 Oct 2004
Posts: 215
Location: Basingstoke, UK
|
Just a quicky. I know what the -m part represents but cant remember what the -fip is. I seem to to have hundreds of the things for a small number of applications. Still I have close to 150 for the amqrmppa process which seems a lot for 20 channels.
Can someone tell me what the fip represents and why so many. Cheers
Solaris running
Name: WebSphere MQ
Version: 530.5 CSD05
CMVC level: p530-05-L030926
BuildType: IKAP - (Production)
_________________
Keith Guttridge
-----------------
Using MQ since 1995 |
|
| Back to top |
|
 |
| JasonE |
| Posted: Wed Nov 17, 2004 2:01 am Post subject: |
|
|
Grand Master
Joined: 03 Nov 2003
Posts: 1220
Location: Hursley
|
Can you please explain this in more detail as I dont fully understand your question.
You should have an amqzlaa0 thread per mqconnected (non-fastbound) app. Similarly you should have an amqrmppa thread for each active channel, which in turn should connect. What exactly are you seeing? |
|
| Back to top |
|
|
| KeeferG |
| Posted: Wed Nov 17, 2004 2:11 am Post subject: |
|
|
Master
Joined: 15 Oct 2004
Posts: 215
Location: Basingstoke, UK
|
Morning JasonE
What we are seeing is the system spawning more and more amqzlaa0 anq amqrmppa processes spawning until there are no more proceses left. I am trying to figure out what is causing it. when doing a ps -ef the amqzlaa0 process has a -m for the queue manager and a -fip with a number. I was wondering what the fip relates to. The system is a cluster of 16 queue managers, each talking to every other one so the most channel we should have is 15 clussdr and clusrcvr per box.
The last box that died had over 500 amqrmppa and 300 amqzlaa0 processes.
We think its something to do with queues filling up and channels going into paused status and causing problems.
_________________
Keith Guttridge
-----------------
Using MQ since 1995 |
|
| Back to top |
|
|
| JasonE |
| Posted: Wed Nov 17, 2004 2:23 am Post subject: |
|
|
Grand Master
Joined: 03 Nov 2003
Posts: 1220
Location: Hursley
|
If you are seeing more and more amqzlaa0 and amqrmppa processes, at a 1:1 ratio, then you effectively have more and more channels running. Do you have clients connecting to the box? This would be my first guess - clients not disconnecting. I'm suprised by the 500 rmppa's and 300 zlaa0's - are you talking threads or processes (Can you count threads/process, are you using multithreaded agents etc?)
What does dis chs(*) show - multiple entries for any channels? |
|
| Back to top |
|
|
| KeeferG |
| Posted: Wed Nov 17, 2004 3:01 am Post subject: |
|
|
Master
Joined: 15 Oct 2004
Posts: 215
Location: Basingstoke, UK
|
There are no client connections in the system. When the system is healthy we only ever have 1 amqrmppa and 5 amqzlaa0 processes using ps -ef | grep amq.
Once queues fill up then we sometimes see this issue. It is more prevelent when we have a low HBINT on the channels.It is almost as if the HBINT is sent while the receiver is in PAUSED state and so spawns a new chanel as the receiver MCA is busy trying to put the message.
Here are the channel defs now
dis channel(TO.*) all
1 : dis channel(TO.*) all
AMQ8414: Display Channel details.
CHANNEL(TO.SP.AR01) CHLTYPE(CLUSSDR)
TRPTYPE(TCP) DESCR(Cluster-sndr channel)
MCANAME( ) MODENAME( )
TPNAME( ) BATCHSZ(50)
DISCINT(0) SHORTRTY(10)
SHORTTMR(5) LONGRTY(999999999)
LONGTMR(10) SCYEXIT( )
SEQWRAP(999999999) MAXMSGL(4194304)
CONVERT(NO) SCYDATA( )
USERID( ) PASSWORD( )
MCATYPE(THREAD) CONNAME(10.252.18.21(1420))
HBINT(20) BATCHINT(0)
NPMSPEED(NORMAL) SSLCIPH( )
BATCHHB(20) LOCLADDR( )
KAINT(AUTO) MCAUSER( )
CLUSTER(CLUSP) CLUSNL( )
ALTDATE(2004-11-13) ALTTIME(19.25.15)
SSLPEER()
MSGEXIT( )
SENDEXIT( )
RCVEXIT( )
MSGDATA( )
SENDDATA( )
RCVDATA( )
AMQ8414: Display Channel details.
CHANNEL(TO.SP.AS03) CHLTYPE(CLUSRCVR)
TRPTYPE(TCP) DESCR(Cluster-receiver channel)
MCANAME( ) MODENAME( )
TPNAME( ) BATCHSZ(50)
DISCINT(0) SHORTRTY(10)
SHORTTMR(5) LONGRTY(999999999)
LONGTMR(10) SCYEXIT( )
SEQWRAP(999999999) MAXMSGL(4194304)
PUTAUT(DEF) CONVERT(NO)
SCYDATA( ) MCATYPE(THREAD)
CONNAME(10.252.18.33(1420)) MREXIT( )
MRDATA( ) MRRTY(10)
MRTMR(1000) HBINT(20)
BATCHINT(0) NPMSPEED(NORMAL)
NETPRTY(0) SSLCIPH( )
SSLCAUTH(REQUIRED) BATCHHB(20)
LOCLADDR( ) KAINT(AUTO)
MCAUSER( ) CLUSTER(CLUSP)
CLUSNL( ) ALTDATE(2004-11-13)
ALTTIME(19.25.15) SSLPEER()
MSGEXIT( )
SENDEXIT( )
RCVEXIT( )
MSGDATA( )
SENDDATA( )
RCVDATA( )
We found that HBINT of 5 caused lots of problems. HBINT of 10 was less frequent and it has only happend with the HBINT of 20
_________________
Keith Guttridge
-----------------
Using MQ since 1995 |
|
| Back to top |
|
|
| FrankBo |
| Posted: Mon Jul 24, 2006 8:23 am Post subject: amqzlaa0 process and fip |
|
|
Apprentice
Joined: 08 Dec 2005
Posts: 36
|
I've read this post and would like to get your info as to why the AMQZLAA0 jobs begin and end (other than when the sbs is ended/started)? Is there a simple explanation?
Thanks |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Mon Jul 24, 2006 2:22 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
Read up in the intercommunications manual on HBINT. I believe there is a precise definition there and it tells you why it is bad to have it too low.
Enjoy 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| JasonE |
| Posted: Tue Jul 25, 2006 1:37 am Post subject: |
|
|
Grand Master
Joined: 03 Nov 2003
Posts: 1220
Location: Hursley
|
| Well the simple approximate answer to the amqzlaa0 processis is that there is one thread in an amqzlaa0 process which is responsible for each MQCONN connection handle. To avoid one process with all the threads in, they are limited, so after 64 (default), we launch another process to handle the next one. When a MQDISC disconnect occurs, the thread is finished with and will (roughly, ignoring any cache/wait for reuse) end. When all threads are gone, the process can end. |
|
| Back to top |
|
|
|
|
