RE: E-business Concmgr goes down due to database crash

From: Mark W. Farnham <mwf_at_rsiz.com>
Date: Mon, 1 Aug 2011 15:00:01 -0400
Message-ID: <018f01cc507d$38606fc0$a9214f40$_at_rsiz.com>


  1. If you don't immediately restart the concurrent managers you can look at the status in requests. Of course the status is running at that point. It shouldn't be too difficult to query the requests to organize emails by user for jobs in this state BEFORE you restart the concurrent managers.
  2. Jobs that do a huge amount of work between commits or with a single endpoint commit may result in a long rollback before they can be restarted. You can get a rough idea (proportional - rollbacks are usually faster than the original transaction unless your undo is functionally unusable beyond all recognition) from the difference between the actual start time of the job and the crash (for single commit point jobs). You want to think seriously about how much rollback has to take place before you turn on the concurrent managers and let users loose on the system.

good luck. I bet a bunch of folks send in canned queries.  

mwf  

From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Sanjeev M
Sent: Monday, August 01, 2011 2:12 PM
To: ORACLE-L
Subject: E-business Concmgr goes down due to database crash  

Gurus,  

We have oracle e-business environment and when concurrent manager goes either due to database or concmgr host outage, my understanding is whatever jobs that were running at that time of failure get erorred and when outage is over and when we(or FNDREVIVER) start concurrent manager it picks up jobs based on their next schedule run time. I am just wondering if you guys at any time had to inform users on what jobs have failed when ConcMgr went down? Do you have any useful queries( on fnd_concurrent_requests) I am thinking on getting all requests between outage window irrespective of their phase/status code(As db session might not have got a chance to update it when db crash abruptly during outage window). Our mgmt wants us to send this email to users about their job failures. I guess there should be better way for users to check this also. Please let me know if any of the apps dba's out there are doing this.  

Regards,

Sanjeev.

--
http://www.freelists.org/webpage/oracle-l
Received on Mon Aug 01 2011 - 14:00:01 CDT

Original text of this message