Edit: Hopefully none of you are running this version these days. Updated the link to Chris’ blog. Updated link to Chris’ blog post. Only one of the POWER7 links at the end still works.
Originally posted January 26, 2010 on AIXchange
Chris Gibson, who blogs about AIX at IBM developerWorks, recently wrote about some issues he encountered while trying to install VIOS fixpack 2.1.2.10-FP-22.
I recently had similar issues with this VIOS fixpack. When I ran the updateios command, I still had virtual media attached from the vio server to a vio client. When I tried to use my virtual media repository, I received this message:
$ lsrep
unable to retrieve repository date due to incomplete repository structure
I initially wondered if VIOS no longer supported the direct copying of .iso files into the /var/vio/VMLibrary, and if I should have used the mkvopt command instead. As Chris mentions, I also found that running the rmrep and mkrep commands fixed the problem with my virtual media repository.
I had called IBM Support around this time, as I was having other issues loading NetApp filesets into my vio server after running oem_setup_env.
I would get errors:
503-466 installp: The build date requisite check failed for fileset ios.cli.rte. Installed fileset build date is 0920. Selected fileset does not have a build date, but one is required. installp: Installation failed dueto BUILDDATE requisite failure.
We also found lppchk –v errors:
$ioslevel
2.1.2.10-FP-22
#oslevel -r
6100-03
# lppchk -v
lppchk: The following filesets need to be installed or corrected to bring
the system to a consistent state:
ios.cli.rte 6.1.4.1 (usr: COMMITTED, root: not installed)
At the time Support had me load an efix onto the VIO server, but that’s since been superseded by level VIOS 2.1.2.10-FP22.1.
This just highlights the need to load fixes in a test environment before rolling them straight into production. Of course, some shops lack the necessary test and QA gear for this. Other shops put off upgrades, figuring that if it isn’t broken, don’t mess with it. And anyone can test upgrades and still miss subtle bugs that only appear under certain conditions.
Most people recommend doing as much testing as possible before making changes to production systems. Backing out of a change to production data is usually non-trivial, particularly if the systems are running for several days before a problem is found. At this point you can’t just go to a backup tape; you must also decide what to do with all of the changed data. No administrator wants to see an issue arise and have to wait for a fix to be created and applied to the system.
So what’s your approach to testing? Do you typically take the latest and greatest OS code and fixes and apply it in production, or do you usually lag a bit and make sure that things work in your test and QA environments before running them in production?
On another note, I’m seeing more information about POWER7. Check out this and this.