Bug #12116
closed
Added by Zack Cerza almost 9 years ago.
Updated about 8 years ago.
Description
teuthology-nuke is failing on lots of systems because of:
[ubuntu@magna069 ~]$ sudo rpm -ev ceph
Preparing packages...
error reading information on service ceph: Invalid argument
error: %preun(ceph-0.80.9-255.gd0f9c5f.el7.x86_64) scriptlet failed, exit status 1
error: ceph-0.80.9-255.gd0f9c5f.el7.x86_64: erase failed
- Assignee set to Ken Dreyer
- Status changed from New to In Progress
I can't reproduce this by doing a simple "yum install ceph; rpm -ev ceph" on the tip of firefly on magna069 (the package uninstalls cleanly).
My guess is that the conditions that trigger this bug are more subtle. Perhaps the fact that teuthology-nuke runs killall on all the Ceph procs prior to uninstalling the package, leaving the service in some inconsistent state? Maybe the servers have to be running a real cluster? It's hard to tell where "invalid argument" comes from without further information (like running the init script with -x or something to get more information)
In any case, I think it's safe to simply ignore the error code of the "service stop" command during uninstall. https://github.com/ceph/ceph-ci/tree/wip-12116-rpm-preun does this.
- Status changed from In Progress to Can't reproduce
Also available in: Atom
PDF