Ok, I attempted to apply a patch to a V440 and it bailed part way through. When I tried to patchrm the patch (after booting from CD), patchrm said that I had to fully install the failed patch before I could delete it. Since installing it killed the system forcing a boot from CD, it sounds like a catch-22. Can't install due to failure, can't delete due to incomplete install.
It's possible I could pkgadd the three undo files (/var/sadm/pkg/SUNW[csl|csr|csrx]/save/109147-30/undo.Z) except that the -R option doesn't seem to work as I expect it to.
I get a "can't update /var/sadm/..." some file which of course is true since / is the mounted boot CD.
Thoughts? Ideas on where to go forward on this? I'm about to the "reinstall the OS and let apps reinstall; they have backups, right?" point but would still like to recover without having to take a bunch of time reconfiguring the system.
I'm not sure what to do. However, when doing a bit of checking I did find on thing. The patchadd manpage (at least for Sol 10) has a very long list at the end showing the possible error messages as well as workarounds. If you haven't looked at that yet, see if you can find the error message you got when it first failed on that list. Maybe you'll find something useful there.
Well patchadd was the tool that I used to attempt to install the patch. patchadd calls pkgadd to install a minimal version of the package that's being patched. As part of the backout, the old files are stored as a compressed datastream package in /var/sadm/pkg/SUNWcsl/save/109147-39/undo.Z. The uncompressed undo.Z file is the datastream file.
I was thinking I could pkgadd that undo file and recover from the bad patch but I wasn't able to get pkgadd -R /mnt [path.to.pkg] working. It was trying to update a /var/sadm/... file and that's the mounted CD.
patchrm failed because 109147-39 wasn't completely installed and I couldn't fully install 109147-39 because patchadd -R /mnt [patch] was failing from CD. Again for unknown reasons.
catch-22.
That's why I was going to attempt to manually apply the datastream packages.
I spoke to Sun support yesterday and they admitted there was a problem with 109147-39 but couldn't pin down if our configuration was affected by the problem (the Sun second level support guys weren't familiar with the V440's which didn't help).
Ultimately we restalled the OS yesterday on both systems. I installed all the necessary patches last night excluding 109147-40 (released Monday). Then, with crossed fingers and toes, I installed 109147-40 without issue.
The main problem was how to recover when the tool used to recover wasn't working as expected.
Heh, believe me, I went through the man pages with a fine toothed comb as well as searching sunforums, this site, linuxquestions (both) and the web in several places. I found some interesting tools to fix little additional problems but nothing that addressed this one in particular.