View previous topic :: View next topic
|
Author |
Message |
Nagendran
New User
Joined: 24 Jan 2007 Posts: 89 Location: USA
|
|
|
|
Hi,
An IMS job got abended with U0777(deadlock)
Could anybody tell me how to view IMS log to identify the reason for deadlock?
thanks,
Nagendran.R |
|
Back to top |
|
|
Sinkaravelan S
New User
Joined: 21 Apr 2008 Posts: 27 Location: coimbatore-TN india
|
|
Back to top |
|
|
udit_adhlakha
New User
Joined: 19 Feb 2008 Posts: 5 Location: India
|
|
|
|
I prepared this doc during my project and this is related to my project. Please see if this comes handy :
Steps to follow while doing RCA for Contention Issues:-
• The job name and ABEND code would be mentioned in JHS. Also we need to know the step which caused the ABEND. This will help us to track the exact position of the ABEND.
• The data for the abend is usually present in JHS or the JHS logs generation dataset - please check where the logs are stored.
• We need to locate the exact position of our abend in JHS/log file. We can search with the job name, ABEND code and/or step name (in which the job abended) in the logs and we would find the contention information there.
• A common form of the contention information mentioned would be like this :
DSNT375I -DB2A PLAN=DDBPGSC0 WITH 985
CORRELATION-ID=0033DDBPGSC0
CONNECTION-ID=IMSJ
LUW-ID=CDN.DHIPA040.C20AFF550A65=134682
THREAD-INFO=J081TCG:*:*:*
IS DEADLOCKED WITH PLAN=DDBCLDV0 WITH
CORRELATION-ID=0041DDBCLDV0
CONNECTION-ID=IMSJ
LUW-ID=CDN.DHIPA040.C20AFF7A93FA=134856
THREAD-INFO=J081TCG:*:*:*
ON MEMBER DB2A
The above means, the job that abended in the step which was executing the program ‘DDBPGSC0’ (this can be verified in the JCL) and it had a contention with the program ‘DDBCLDV0’.
Search for the job that uses the second program and that would be the job, the abended job had a contention with.
Sometimes the job which uses the above mentioned program would also be mentioned. Basically the contention information mentioned in the logs can take many forms but we can find the information in the logs. |
|
Back to top |
|
|
Ajay Baghel
Active User
Joined: 25 Apr 2007 Posts: 206 Location: Bangalore
|
|
|
|
Hi friends,
We are facing U0777 abend frequently during our batch. One job is there that runs for 7-8 hours and it may be the reason for contention with other 10 other jobs that run during this 7-8 hours window since all these jobs use the same database.
Both this job and any of those 10 jobs keep on dying because of this deadlock. Often i don't know which one to restart.
I was wondering, if we could do a quick lookup while fixing the abend to see who is HOLDING the resource and who is WAITING for the resource and what is that type of resource.
Hi Udit,
We are getting U0777 abend but not one containing DEADLOCK keyword.
For eg:
DSN3201I -DB2A ABNORMAL EOT IN PROGRESS FOR 038
USER=GCAXQ CONNECTION-ID=G2P2 CORRELATION-ID-0109TQAAMEE4
JOBNAME=TQAAMEE4 ASID=01C3 TCB=009FF050
IEF450I TQAAMEE4 GO ST3ABAS - ABEND=S000 U0777 REASON=00000000 039
Please help.
Thanks,
Ajay |
|
Back to top |
|
|
Ajay Baghel
Active User
Joined: 25 Apr 2007 Posts: 206 Location: Bangalore
|
|
|
|
No voice heard till now... So quiet. hmmm
I wonder if we just need to keep our fingers crossed and restart any job randomly. Is really there any way to get clear about resolving u0777.
Hoping some answers/pointers friends!!
-Ajay |
|
Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
|
|
|
|
Hello,
Usually, deadlocks are due to design problems. There is no "magic" to make them go away. It usually takes considerable effort.
One way around the design problems is scheduling.
Another is to significantly reduce the runtimes of the problem job(s) - in your case probably the 7-8 hour job.
Lastly, the most effective but the most work, correct the design to prevent the deadlocks. |
|
Back to top |
|
|
Ajay Baghel
Active User
Joined: 25 Apr 2007 Posts: 206 Location: Bangalore
|
|
|
|
Quote: |
Usually, deadlocks are due to design problems. There is no "magic" to make them go away. It usually takes considerable effort.
One way around the design problems is scheduling. |
Quote: |
Lastly, the most effective but the most work, correct the design to prevent the deadlocks. |
You are right Dick. This is obviously a design problem that needs a serious makeover.
But for doing schedule changes, we need to identify correctly the jobs involved in the deadlock at a given moment. How do we do that? That is my question.
While doing google search, i got some information on generating Deadlock reports and DFSERA utilities.
Continuing with my quest gracefully and hopefully ...
-Ajay |
|
Back to top |
|
|
Ajay Baghel
Active User
Joined: 25 Apr 2007 Posts: 206 Location: Bangalore
|
|
|
|
Hi Friends,
I have got what i was looking for. We can generate human-readable deadlock reports using DFSERA10 and DFSERA30 utilities.
Thanks,
Ajay |
|
Back to top |
|
|
|