udev not using mdadm incremental

[Daniel Miller]

Problem occurs when activating an existing raid array on boot. My configuration, which exposed this, was four SATA drives in a Raid-10 configuration created with mdadm. On boot, udev appears to be entering some race conditions with mdadm - the array is activated with only three of the drives.

This appears to be an issue with using the "auto" detection of mdadm. Consulting the linux-raid mailing list shows the "incremental" flag was expressly created for udev - but it doesn't appear to have be being used by udev! Please contact the mdadm developers (like Neil Brown) to get this resolved.

[Scott James Remnant (Canonical) (canonical-scott)]

Can you supply a reference to the mailing list post defining --incremental

I suspect this option has been added since our mdadm packages were frozen.

[Daniel Miller (dmiller)]

Here's a link to the reply I received from Neil Brown:

http://<email address hidden>/msg09552.html

Let me know if you need more from me.

[ceg (ceg)]

--incremental is available in hardy (8.04)

Some info on howto enable it is available on:
https://wiki.ubuntu.com/BootDegradedRaid

[ceg (ceg)]

When this occurs the drives are accessed continously, system hangs.

[ceg (ceg)]

Sorry prior post was ment for another bug, the race condition with degraded raids in this bug has different behaviour mentioned in aformentioned kernle list archive.

[xteejx (xteejx-deactivatedaccount)]

Hi guys,
Thanks for reporting this bug to us. Do you know if this happens to affect Karmic? I do not have any kind of RAID setup to be able to test this. Thank you.

[ceg (ceg)]

Hi,
In 9.10 someone changed the mdadm udev rules to use --incremental without considering all the consequences that were already reported. https://wiki.ubuntu.com/ReliableRaid

[SoulWax (nicolas-melchior)]

Hello,

There is always the same problem:

after creating a raid mirror named md0 by example, it works until the reboot of the system

after the rebbot, I only have a /dev/md_d0 wich is set unactive with only one of my drive .

I have to stop it, and do mdadm --assemble --scan (or --auto-detect) for having my raid fonctionning again. So I think it's coming from udev as mdadm alone works fine to re-detect the /dev/md0 drive.

[ceg (ceg)]

@SoulWax: Did you follow and use the manual fix of the ubuntuforums link on https://wiki.ubuntu.com/ReliableRaid ?

Sounds like bug #252345