Change in vdsm[master]: virt: Force a vdsm restart on too many epoll errors on untra...

michal.skrivanek at redhat.com michal.skrivanek at redhat.com
Thu Jun 25 09:24:06 UTC 2015


Michal Skrivanek has posted comments on this change.

Change subject: virt: Force a vdsm restart on too many epoll errors on untracked FDs
......................................................................


Patch Set 3:

(1 comment)

https://gerrit.ovirt.org/#/c/42569/3/vdsm/virt/vmchannels.py
File vdsm/virt/vmchannels.py:

Line 182:             self._handle_event(fileno, event)
Line 183:             if self._untrackedErrors > UNTRACKED_ERROR_RESTART_LIMIT:
Line 184:                 self.log.critical("Too many epoll errors happened."
Line 185:                                   "Taking down process")
Line 186:                 os.kill(os.getpid(), signal.SIGTERM)
> This scares me. I understand that if python process losts too much terrain,
one thing to consider - we used to stop only the ga listener thread on issues, leaving the rest of vdsm running. And the experience is even worse, you think it works...but actually it doesn't, vital information is not being reported (e.g. balloon, or IPs) crippling the functionality later...and it's difficult to correlate those events. 
So +1 for termination

whether we can get into a loop...hard to say...it probably may happen when qemu process goes wild, but what can we do. perhaps better quarantine such VM? Maybe run an epoll test on recovery?
Line 187:         else:
Line 188:             self._update_channels()
Line 189:             if (self._timeout is not None) and (self._timeout > 0):
Line 190:                 self._handle_timeouts()


-- 
To view, visit https://gerrit.ovirt.org/42569
To unsubscribe, visit https://gerrit.ovirt.org/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I654e5b828e2c05e8a86fa0e945be67cec014c110
Gerrit-PatchSet: 3
Gerrit-Project: vdsm
Gerrit-Branch: master
Gerrit-Owner: Vinzenz Feenstra <vfeenstr at redhat.com>
Gerrit-Reviewer: Dan Kenigsberg <danken at redhat.com>
Gerrit-Reviewer: Francesco Romani <fromani at redhat.com>
Gerrit-Reviewer: Gal Hammer <ghammer%redhat.com at gtempaccount.com>
Gerrit-Reviewer: Jenkins CI
Gerrit-Reviewer: Michal Skrivanek <michal.skrivanek at redhat.com>
Gerrit-Reviewer: Michal Skrivanek <mskrivan at redhat.com>
Gerrit-Reviewer: Vinzenz Feenstra <vfeenstr at redhat.com>
Gerrit-Reviewer: automation at ovirt.org
Gerrit-HasComments: Yes


More information about the vdsm-patches mailing list