https://bugzilla.novell.com/show_bug.cgi?id=731230 https://bugzilla.novell.com/show_bug.cgi?id=731230#c10 --- Comment #10 from Craig Rogers <rogers@isi.edu> 2011-11-28 23:07:16 UTC --- Thinking about this problem some more, what's really desired, and what systemd should be designed to support, is a finer-grained system that tracks dependencies on a per-filesystem basis. For example, ordinary local filesystems could be mounted in parallel with each other, and in parallel with the mdadm instance that is building the RAID devices (/dev/md0, etc.). Once a new raid device is available, an event of some sort would be generated so systemd could then attempt to mount the filesystem on the new device. This could be viewed as an extension of mount point ordering -- there might be filesystems that with mount points that lie on top of the RAID-based filesystem. It might also provide a better system administration experience in case of partial drive or filesystem failure. A compromise might be to split localfs mounts into two phases: one that mounts filesystems that don't depend upon md drives, and one that mounts filesystems that do depend on md drives. Ultimately, I imagine, all this complexity will be hidden in the filesystems themselves, such as with the RAID support integrated into BTRFS. In the short term, I'd be satisfied to have openSUSE 12.1's systemd-based initialization build all the md devices, then mount the local filesystems, just like the SysV init package does. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.