| Author |
Message
|
| KCampbell |
 Posted: Thu Jun 17, 2004 6:38 pm Post subject: ODBC connections won't reconnect after outage? |
 |
|
Newbie
Joined: 17 Jun 2004
Posts: 2
Location: Seattle
|
We've noticed, originally under 2.1, that the ODBC connections established the first time a flow is executed that includes database nodes, will not reconnect if the target database goes down and then reappears.
Is this behavior that we can control? Currently our only option appears to be restarting the broker.
Kevin |
|
| Back to top |
|
 |
| kirani |
| Posted: Thu Jun 17, 2004 10:37 pm Post subject: |
|
|
Jedi Knight
Joined: 05 Sep 2001
Posts: 3779
Location: Torrance, CA, USA
|
Yes, the message flow can re-establish the connection for the second time. You need to increase your Backout threshold propery of the input queue so that message can retry processing. This has been discussed in this forum for multiple times. Please search this form for ODBC and you might find the related post.
_________________
Kiran
IBM Cert. Solution Designer & System Administrator - WBIMB V5
IBM Cert. Solutions Expert - WMQI
IBM Cert. Specialist - WMQI, MQSeries
IBM Cert. Developer - MQSeries
|
|
| Back to top |
|
|
| maxis |
| Posted: Fri Jun 18, 2004 12:35 am Post subject: Re: ODBC connections won't reconnect after outage? |
|
|
Centurion
Joined: 25 Jun 2002
Posts: 144
|
| KCampbell wrote: |
Is this behavior that we can control? Currently our only option appears to be restarting the broker.
Kevin |
you dont necessarily have to restart the broker. Even restart of exe-group ( ie. kill it .. and it'll be RE-STARTED ) will connect to db |
|
| Back to top |
|
|
| kirani |
| Posted: Fri Jun 18, 2004 10:35 am Post subject: |
|
|
Jedi Knight
Joined: 05 Sep 2001
Posts: 3779
Location: Torrance, CA, USA
|
Beware of other side effects that might occur when planning on killing the EG. It'd be a good recovery test though to see how your message flows behaves!! The key point is to build robust error handler that will take care of similar problems.
_________________
Kiran
IBM Cert. Solution Designer & System Administrator - WBIMB V5
IBM Cert. Solutions Expert - WMQI
IBM Cert. Specialist - WMQI, MQSeries
IBM Cert. Developer - MQSeries
|
|
| Back to top |
|
|
| JohnMetcalfe |
| Posted: Wed Feb 02, 2005 8:29 am Post subject: |
|
|
Apprentice
Joined: 02 Apr 2004
Posts: 40
Location: Edinburgh, Scotland
|
Can anyone confirm that this problem is fixed in WBIMB 5. We are on WMQI 2.1, and raised a PMR about this issue - their response was that this was a know problem and was fixed in 2. 1CSD6 and WBIMB 5, fix id IY37803.
As stated by IBM the problem is :
| Quote: |
Broker fails to reconnect to the Broker database after a communications failure.
The DFE does not re-connect to the Broker database if database connection is lost (i.e. database has been stopped/re-started).
The next time that the DataFlowEngine needs to access the database (e.g. during a deploy) it fails and the db connection
is released. |
|
|
| Back to top |
|
|
| JT |
| Posted: Wed Feb 02, 2005 9:03 am Post subject: |
|
|
Padawan
Joined: 27 Mar 2003
Posts: 1564
Location: Hartford, CT.
|
| I don't know about the broker database, but as of v5.0.2, we still receive ORA-03113 errors ('end-of-file on communication channel') for our user databases. |
|
| Back to top |
|
|
| PGoodhart |
| Posted: Wed Feb 02, 2005 9:24 am Post subject: |
|
|
Master
Joined: 17 Jun 2004
Posts: 278
Location: Harrisburg PA
|
The problem still exists in V5. The problem is "fixed" as far as the broker databases but the client databases still have the connection closed issue.
_________________
Patrick Goodhart
MQ Admin/Web Developer/Consultant
WebSphere Application Server Admin |
|
| Back to top |
|
|
| Heinz57 |
| Posted: Wed Feb 02, 2005 11:46 am Post subject: |
|
|
Apprentice
Joined: 21 Jan 2004
Posts: 26
Location: Syracuse, NY
|
| When we brought this issue to IBM, their reply was that it was working as designed. While multiple attempts (based on backout threshold) MAY reconnect, our design is to run MQSIRELOAD when we lose a DBMS. this stops/starts all of the dfe's and fresh db threads are established. This relates to application/client databases NOT broker or configmgr databases. |
|
| Back to top |
|
|
| JLRowe |
| Posted: Wed Feb 02, 2005 11:51 am Post subject: |
|
|
Yatiri
Joined: 25 May 2002
Posts: 664
Location: South East London
|
| I have noticed that if an ODBC connection goes bad because of some kind of error, the bad connection is put back into the connection pool. I noticed this with testing SQL statements on oracle, if I get an SQL error on one message, successive messages get that bad ODBC connection and continue to fail. |
|
| Back to top |
|
|
| shra_k |
| Posted: Wed Feb 02, 2005 9:52 pm Post subject: |
|
|
Apprentice
Joined: 30 Dec 2002
Posts: 37
|
This reconnection problem is fixed and fix is available in V5.0.4.
I have tested this and found things to work normal. My broker is on windows while the DB (db2) is on AIX. |
|
| Back to top |
|
|
| damianharvey |
| Posted: Wed Sep 28, 2005 2:21 pm Post subject: |
|
|
Acolyte
Joined: 05 Aug 2003
Posts: 59
Location: Sydney, Australia
|
| shra_k wrote: |
This reconnection problem is fixed and fix is available in V5.0.4.
I have tested this and found things to work normal. My broker is on windows while the DB (db2) is on AIX. |
I had this today running CSD5. |
|
| Back to top |
|
|
| sieijish |
| Posted: Thu Sep 29, 2005 1:47 am Post subject: |
|
|
Acolyte
Joined: 29 Nov 2004
Posts: 67
Location: London
|
I am also running CSD05 and had noticed the same behaviour. I am calling an oracle stored procedure from my esql. If i get the following errors, it won't resolve itself.
- 'end-of-file on communication channel'
- 'esql deployed is invalid' happens when broker can't connect to oracle to validate my external database procedure.
But I am not able to consistantly replicate the problem as sometimes I have seen it working fine for the second message. |
|
| Back to top |
|
|
| damianharvey |
| Posted: Thu Sep 29, 2005 1:01 pm Post subject: |
|
|
Acolyte
Joined: 05 Aug 2003
Posts: 59
Location: Sydney, Australia
|
Turns out the Oracle DBA's rebooted the Database Servers but neglected to tell us. We had 3 flows fail. But each flow only had the first record fail. After the first record everything processed successfully. This would indicate that the Message Flows had stale connections that were refreshed when the first connection failed.
I will be able to get around this by implementing a retry loop in the Flow. |
|
| Back to top |
|
|
| whbrownnc |
| Posted: Wed Nov 16, 2005 3:39 pm Post subject: Broker DB connections question on mqsireload |
|
|
Novice
Joined: 07 May 2003
Posts: 20
Location: Charlotte, NC
|
We are planning to implement the use of "mqsireload" to enable an execution group to recover from a lost application DB outage.
Does the mqsireload release broker database resources cleanly? Does it have any untoward "side effects" such as leaving behind outstanding semaphores or memory segments? Will DB2 "complain"?
(Broker is running at v5 CSD04 on AIX 5.2 ML10, DB2 v8.1.1.80, WMQ 5.3.0.
_________________
Thanks,
Bill Brown
Integration Hosting and Services
IBM Certified - WMQ 6.0, 2008
IBM Certified - WMB 6.0, 2008 |
|
| Back to top |
|
|
| jbanoop |
| Posted: Mon Nov 13, 2006 12:46 am Post subject: |
|
|
Chevalier
Joined: 17 Sep 2005
Posts: 401
Location: SC
|
Didnt think it was necessary to raise a new thread for this ..
a quick question ..
Is this issue with stale connections causing the flow to rollback for the first time addressed in V 6 ?
Regards,
Anoop |
|
| Back to top |
|
|
|
|
