Mailinglist Archive: opensuse-bugs (5243 mails)

< Previous Next >
[Bug 793954] raid1 array sometimes started with one of two disks at boot (again)

https://bugzilla.novell.com/show_bug.cgi?id=793954

https://bugzilla.novell.com/show_bug.cgi?id=793954#c41


--- Comment #41 from Neil Brown <nfbrown@xxxxxxxx> 2013-03-21 03:22:15 UTC ---
I agree that systemd doesn't have a problem. The documentation for ExecCommand
doesn't meantion ';', but the doco for ExecCommandStart does, and I found where
the code makes a special case of ';' so that is all good.

I may have been a bit hasty is combining all these bugs together but they do
look very similar. The core problem is that udev is running "mdadm -I" for
various devices *after* "udevadm settle" has completed.

This can definitely happen if the "udev-trigger" systemd service starts after
"udevadm settle" was called, so making boot.md depend on udev-trigger can close
that hole - and it definitely needs to be closed for 12.3.

But it seems that there is another hole, and I cannot find it. And as I cannot
reproduce it, that makes it extra hard.

Maybe if you could capture an 'strace' of the "udevadm settle" call made by
boot.md. That might help - and it might help even for a case where no problem
with raid arrays is seen, though a cas where the problem does occur would be
better.

i.e. modify /etc/init.d/boot.md so that the "udevadm settle --timeout=...."
line is
strace -o /tmp/udevadm.trace -s 300 udevadm settle --timeout=....

i.e. just put "strace -o /tmp/udevadm.trace -s 300" in front of the command.
If you can do that, then attach the /tmp/udevadm.trace file it might help.

--
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.

< Previous Next >