This happened when I tried to switch the SD "mi-ci-3" to maintenance mode while copying template's disk image from "mi-ci-1" to "mi-ci-5" (and "mi-ci-3" also had this image already copied). I have already fixed this by some changes in DB and destroy SD. My installation is not self-hosted deploy.

I'll try to dig logs from that situation today and post it here (I produce really a loot of logs)...


2015-04-13 17:45 GMT+02:00 Bob Doolittle <bob@doolittle.us.com>:
Are you using a self-hosted deployment? If so, this may be the issue that's been reported for almost a year now that results in an inability to shutdown cleanly due to sanlock leases being held which result in watchdog timeouts that cause a system to reboot instead of shutdown. See for example: http://lists.ovirt.org/pipermail/users/2014-May/024652.html

I still experience this with 3.5.1. When I poweroff my machine I have to physically be there to hit the power switch when it tries to reboot after the wdmd watchdog timeout. I notice during shutdown that the storage domains fail to unmount.

-Bob