| Author |
Message
|
| jmac |
 Posted: Fri Apr 02, 2004 1:50 pm Post subject: Hold Queue messages |
 |
|
Jedi Knight
Joined: 27 Jun 2001
Posts: 3081
Location: EmeriCon, LLC
|
I have encountered a system that in the span of a little over an hour put 85 messages on the hold queue. In that period approximately 7800 instances were started. The system is MQWF 3.4.0.3 running unix. with a DB2 database There were many database deadlocks in the system log. At this point the database is being reorged, I am mainly asking if anyone else has ever seen an accumulation of messages on the hold queue in that short span of time. If you have what did you find to be the cause.
Thanks
_________________
John McDonald
RETIRED |
|
| Back to top |
|
 |
| koko |
| Posted: Sun Apr 04, 2004 2:38 pm Post subject: |
|
|
Master
Joined: 26 Sep 2003
Posts: 206
|
Hi Jmac,
This is what I wanted to find out in the post "Workflow Capacity". I have a situation where the external system is pumping upes messages to workflow and starting instances. I saw some 38 messages in the exec servers' hold queue.
I would like to know how to *increase* the workflow's capacity to handle the number of messages within a unit of time. I had all the tablespaces filled up and rolled back. The database became a mess.
ENV :workflow 3.4, DB2, 3 tier, AIX
Thanks
_________________
Thanks
Koko |
|
| Back to top |
|
|
| jmac |
| Posted: Sun Apr 04, 2004 2:49 pm Post subject: |
|
|
Jedi Knight
Joined: 27 Jun 2001
Posts: 3081
Location: EmeriCon, LLC
|
I can tell you what has seemed to help my problem:
There were also databse deadlocks, and it seems that when you initally load the DB the tables are of course empty, and that has some affect on the rebind... Not being a database guy, I am just repeating what someone told me. Once I did the runstats & rebind it helped, not sure if it is the final solution, but has definitely lowered the frequency with which items are showing up on the hold queue
_________________
John McDonald
RETIRED |
|
| Back to top |
|
|
| koko |
| Posted: Sun Apr 04, 2004 3:11 pm Post subject: |
|
|
Master
Joined: 26 Sep 2003
Posts: 206
|
Wondering if this is a permanent, good solution. What if the database is Oracle??. As far as I know there is no concept of rebinding for oracle. so any solution which includes "rebinding or re-org" of database should be either specific to DB2 or not a concrete solution. This is just my 0.02Cents.
But I feel there is something that can be done like
increasing the message queue size, using clustering so that load is shared yada..yada...yada...
I would like to know how some folks here are handling this situation.
Thanks
_________________
Thanks
Koko |
|
| Back to top |
|
|
| manoj |
| Posted: Mon Apr 05, 2004 6:22 am Post subject: |
|
|
Master
Joined: 30 Jan 2002
Posts: 237
Location: Virgina
|
john,
how good is your workflow box capacity... i feel like you don't have enough horse power....the workflow server is not able to process all the request in a timely manner , times out and put the messages in to
H-Queue....some people do rebind etc daily and i think i have heard from IBM it's not bad to do daily either.
_________________
-manoj |
|
| Back to top |
|
|
| jmac |
| Posted: Mon Apr 05, 2004 6:37 am Post subject: |
|
|
Jedi Knight
Joined: 27 Jun 2001
Posts: 3081
Location: EmeriCon, LLC
|
manoj:
I will know more later today, but one of the issues appears to have been that there were only 2 ES started on an 8way CPU, that was bumped up to 16, and it appears to have helped, if not solved the problem.
Thanks for your input
_________________
John McDonald
RETIRED |
|
| Back to top |
|
|
| MaheshPN |
| Posted: Thu Apr 08, 2004 9:50 am Post subject: |
|
|
Master
Joined: 21 May 2003
Posts: 245
Location: Charlotte, NC
|
I faced the same problem, when I created 8000 instances at a time.
some were in holdQ and I saw deadlocks on DB too.
I did runstat and rebind as recommended. After that, I did not see any exception.
I want to know, can we do runstat and rebind without bringing down the WF servers? I tried it but I am not sure it is safe and effective.
Thanks,
-Mahesh
IBM Certified Solution Expert |
|
| Back to top |
|
|
| jmac |
| Posted: Thu Apr 08, 2004 9:59 am Post subject: |
|
|
Jedi Knight
Joined: 27 Jun 2001
Posts: 3081
Location: EmeriCon, LLC
|
Mahesh:
The document "Production Workflow considerations" or something like that, which is where this recommendation for the runstats rebind comes from indicates that you must stop the system.
_________________
John McDonald
RETIRED |
|
| Back to top |
|
|
|
|
