Monday, 28 September 2009

If at first you don't succeed, don't try again

I once brought down both the Development and Production mainframes by running the same SQL query on both of them.

When we first started using DB2 the performance was not good enough for online production systems but was OK for decision support.

We had a system where the developers had access to both the development environment and production systems.Since it was a read only system, we could not do much harm; except if we wrote a long running SQL query.

Well one day I ran a very simple query that didn't produce the output instantly, as previously experienced. After a big delay I commented "That's not supposed to happen" and then proceeded to demonstrate to the 'stupid question asker' what was supposed to happen.
"See, it works on production!" I said, rather too soon.

The queries running on both machines were using a shared swapping drive
and trying to write an awful lot of temporary data to the drives.
The only way to fix it was to reboot both machines.

Fortunately for me we were able to blame the DBA for not running runstats after a reorg.

No comments:

Post a Comment