Le jeudi 03 novembre 2011 à 01:02 +0100, Christian Boltz a écrit :
Hello,
Am Mittwoch, 2. November 2011 schrieb Frederic Crozat:
Le mercredi 02 novembre 2011 à 14:36 +0100, Stephan Kulow a écrit :
We still have 2 issues with mysql (#727706 and #712749) and two with systemd (#726823 and #724912). Additionally to that we have
Wow, one of my bugreports is famous ;-)
Right now, #724912 is caused by getting notification from kernel before the raid is active. And unfortunately, I don't see a "smart" way for systemd to detect a mount point is a "md" one and add on the fly a dependency on md.service (ie boot.md) for fsck for this partition.
A not-so-smart and quite obvious idea: check if the device name matches /dev/md* ;-)
Except we don't have a garantee /dev/md* will be the name used in /etc/fstab or /etc/crypttab (it could be /dev/disk/by-...). It would be better to rely on sysfs (/sys/block/virtual/md ) but the change is non trivial and I don't want to break systemd so close to the release.
BTW: Would it hurt to always have "After: md.service" in fsck@.service and cryptsetup@*.service? That would be the easiest solution.
Yes, that is one possibility, but against, I'm a bit afraid of doing
such changes so close to the release. We should probably also add After:
dmraid.service too, I guess.
--
Frederic Crozat