We're slowly working our way through this issue. It occurs when we shut down the underlying vserver host. Despite our creating a /etc/vservers/zimbra1/apps/vshelper/sync-timeout file and setting it to 900 (seconds) to give the system plenty of time to shutdown gracefully, it is not. If we watch the shutdown sequence, it is considering the zimbra server stopped long before it possibly could be. Thus, in effect, we are crashing the zimbra server.
We have escalated to the vserver list to determine how to solve the problem from the vserver side. However, that doesn't mean there will be other reasons why the zimbra server might crash in the future. I'm a bit surprised by how much damage is done and how consistently. Are there settings which can be tweaked to make zimbra a little more resistant to damage during a crash? Thanks - John |