On 03/26/2015 02:51 PM, Peter Krempa wrote:
On Wed, Mar 25, 2015 at 12:19:18 -0400, John Ferlan wrote:
On 03/24/2015 06:06 AM, Erik Skultety wrote:
The checkPool callback should be used when updating states of all pools during storageStateInitialize, not storageDriverAutostart, otherwise we can't start a domain which mounts a volume after daemons restarted. This is because qemuProcessReconnect is called earlier than storageDriverAutostart which marks the pool as active, so that the domain can mount a volume from this pool successfully. In order to fix this, conn attribute has to be discarded from the callback, because storageStateInitialize doesn't have a connection yet. (it's not used anyway)
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1177733
This patch definitely does not resolve this^^.
I'm not convinced we can remove 'conn'. Something about backend or backward compatibility. Since it's not used by any backend anyway, there's (more or less) no harm in keeping it.
The storage driver backend APIs are internal so it's okay to change Additionally I recall that it will simplify some of the further patches.
ACK
Peter
Thanks, I removed the BZ link and reworded the message a bit and pushed, v2 of the remaining patches coming in a second. Erik