IBM Support

Why cold starts are not recommended for WebSphere MQ general recovery

Technote (troubleshooting)


Problem(Abstract)

Why IBM recommends against doing a cold start for WebSphere MQ.

Resolving the problem

If you encounter a problem and use the cold start procedure at V5.3 or later, you might lose messages. If messages are in memory and not yet flushed to disk, they will be lost . The cold start procedure may also damage queue objects. Cold start is NOT a supported procedure. There is no guarantee that a cold start will allow the queue manager to restart to a valid state. There is every possibility that a cold start will impact message integrity. Messages could be lost or gained.


A cold start procedure at earlier releases will lose messages, which may have just been committed, and gain messages, which were just backed out. By doing the cold start you are throwing away the operations which need to be replayed. The difference between V5.2 and V5.3 or later is that the message chains on V5.3 and later may go from a position on the disk, to a position in a memory buffer, and back to a position on the disk. That was never the case in earlier releases.

Any problem which causes the queue manager to not start due to the logs needs to be fixed, not avoided by performing cold starts. In this way MQ can continue to provide our assured message delivery. We can not avoid user errors but any other problems need to be investigated.

Release on release (and sometimes even in Fix Packs or CSDs) more optimizations are put into the product which means the "Q" files are not stand-alone entities from which you can recover. The cold start procedure should not be a regular procedure nor considered a way of recovering from a specific type of problem. In very rare circumstances IBM support might advise a customer to perform a cold start, this is only undertaken as a last resort when a queue manager will not start normally. A cold start can sometimes be preferable to deleting and recreating the queue manager, but there are no guarantees about the integrity of a queue manager recovered in this manner.

Important Note
Whenever you have a queue manager restart failure, then it is essential that you capture a full copy of the queue manager data and recovery logs in order that the restart failure can be thoroughly investigated.

On UNIX systems these are all files under:
/var/mqm/qmgrs/<qmgrname> and
/var/mqm/log/<qmgrname>

On Windows systems these are all files under:
C:\Program Files\IBM\WebSphere MQ\Qmgrs\<qmgrname>
C:\Program Files\IBM\WebSphere MQ\log\<qmgrname>

Product Alias/Synonym

WMQ
WebSphere MQ
MQSeries

Document information

More support for: WebSphere MQ
Recovery

Software version: 6.0, 7.0, 7.0.1, 7.1, 7.5, 8.0

Operating system(s): AIX, HP-UX, Linux, Solaris, Windows

Reference #: 1162238

Modified date: 14 December 2012


Translate this page: