mdadmin can't re-sync a mirror?!

mdadmin can't re-sync a mirror?!

[James Ralston]

Either I've grossly understood how Linux RAID and/or the mdadmin
command works, or this is horribly wrong:

   Device Boot    Start       End    Blocks   Id  System
/dev/sdb1             1        33    265041   fd  Linux raid autodetect

   Device Boot    Start       End    Blocks   Id  System
/dev/sdc1             1        33    265041   fd  Linux raid autodetect

$ mdadm --create /dev/md0 --verbose --level=mirror --raid-devices=2 /dev/sdb1 /dev/sdc1

[wait for mirror to sync]

$ cat /proc/mdstat
Personalities : [raid1]
read_ahead 1024 sectors
md0 : active raid1 sdc1[1] sdb1[0]
      264960 blocks [2/2] [UU]
unused devices: <none>

Okay, now simulate (without physically changing any disks) a failure,
and then a replacement:

$ mdadm /dev/md0 -f /dev/sdc1 -r /dev/sdc1
mdadm: set /dev/sdc1 faulty in /dev/md0
mdadm: hot removed /dev/sdc1

$ mdadm --zero-superblock /dev/sdc1

$ mdadm /dev/md0 -a /dev/sdc1
mdadm: hot added /dev/sdc1

But the mirror doesn't sync (!):

$ cat /proc/mdstat
Personalities : [raid1]
read_ahead 1024 sectors
md0 : active raid1 sdc1[2] sdb1[0]
      264960 blocks [2/1] [U_]
unused devices: <none>

Have I completely misunderstood how Linux software RAID is supposed to
work?  (I certainly hope so, because otherwise, it can't even perform
the simplest of recovery operations: replacing a failed mirror
component and re-syncing.)

Thanks for any comments...

-- 
James Ralston, Information Technology
Software Engineering Institute
Carnegie Mellon University, Pittsburgh, PA, USA

Re: mdadmin can't re-sync a mirror?!

[Friedrich Lobenstock]

Hi!

James Ralston wrote:
> 
> Have I completely misunderstood how Linux software RAID is supposed to
> work?  (I certainly hope so, because otherwise, it can't even perform
> the simplest of recovery operations: replacing a failed mirror
> component and re-syncing.)

Try

   # mdadm --fail /dev/md0 /dev/sdc1
   # mdadm --remove /dev/md0 /dev/sdc1
   # mdadm --zero-superblock /dev/sdc1
   # mdadm --add /dev/md0 /dev/sdc1

This works for me.

-- 
MfG / Regards
Friedrich Lobenstock
____________________________________________________________________
Friedrich Lobenstock         FL226-RIPE             Internetservices
URL: http://www.fl.priv.at/                     Email: fl@fl.priv.at
____________________________________________________________________

Re: mdadmin can't re-sync a mirror?!

[Neil Brown]

On Friday January 24, qralston+ml.linux-raid@andrew.cmu.edu wrote:
> Either I've grossly understood how Linux RAID and/or the mdadmin
> command works, or this is horribly wrong:
...
> 
> But the mirror doesn't sync (!):

Very odd.  What you do should have the desired result, but obviously
doesn't.

You didn't say which kernel you are running - 2.4 or 2.5.xx
I assume there aren't any other RAID arrays present.  Is this correct
(or did you edit the output of /proc/mdstat).

NeilBrown

reproducible kernel Oops in mdrecoveryd (triggered by md raid1 re-syncs)

[James Ralston]

On 2003-01-25 at 15:47:50+1100 Neil Brown <neilb@cse.unsw.edu.au> wrote:

> Very odd.  What you do should have the desired result, but obviously
> doesn't.

Thanks for your response.

Alas, the problem isn't with mdadm.  :(  This is the problem:

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=82815

(I wasn't noticing the Oops in the voluminous output from the md
driver.)

> You didn't say which kernel you are running - 2.4 or 2.5.xx

Red Hat's latest errata kernel for 8.0 (2.4.18-19.8.0).

> I assume there aren't any other RAID arrays present.  Is this
> correct (or did you edit the output of /proc/mdstat).

Nope, no other arrays.

I was looking at the notes for 2.4.21-pre3, and I see there are a
couple of md fixes from you in there.  Unless you have a better idea,
I'll try backporting drivers/md/md.c from 2.4.21-pre3 to Red Hat's
2.4.18-19.8.0 and give that whirl...

-- 
James Ralston, Information Technology
Software Engineering Institute
Carnegie Mellon University, Pittsburgh, PA, USA