Friday, June 19, 2009

IBM PE's seven PTFs for V5R4.

While my pastoral duties at this community are normally rather dull, my favorite activity is working with the most holy code (i5/OS) and PTFs (Parchment Temporary Fixes). I especially love it when I find an error in the code that causes the cardinals and bishops in Rochester to do something drastic.

A couple months ago, as part of a HIPER recommendation, we installed PTF 5722SS1-SI34484. The following Monday morning the BACKUP job was looping forever waiting for an HTTP server job to end. We didn't associate it with the PTF right away because this isn't the first time the backup got held up by something like this. By the time we'd figured out that this was a trend we'd already installed another HIPER recommendation, PTF 5722SS1-SI34576, which superceded the one installed previously.

So we were now stuck in an untenable situation. We couldn't remove the PTF and we didn't want to reload the O/S. Being in this kind of pickle can really focus the mind. I ended up writing a program that detected these zombie HTTP server jobs and killed them with ENDJOBABN.

The minions of our Lord Ib'm eventually discovered that a thread in the HTTP job was prematurely destroying the list of open files. The job would then sit there forever waiting of the files to close not knowing that that list was gone. In the end they had to PE a bunch of PTF's and since two of them were HIPER recommendations the fixing PTF (SI35317) was also made a HIPER.

Final tally was seven PE'ed PTFs and one HIPER, all from one PMR.

No comments:

Post a Comment