split off + mount a partitionless RAID1 device?

split off + mount a partitionless RAID1 device?

[Jeff Breidenbach]

Ok, let's say that a um... hypothetical friend of mine had a power
outage this morning. This many disk RAID1 happily assembled and
mounted (with only mild persuasion) but is now caught up in a
monumentally slow resync. The resync is slow because /dev/md0 is
getting hammered with read requests by a webserver. If I turn off
Apache then resysnc speed shoots up to a respectable level, but I'd
rather not do that.

I'd love to split one of the drives off of the RAID1 and mount it
directly. Then it can feed the ravenous Apache monster while the
other five drives resync in peace. But foolishly and somewhat
accidentally, this RAID is made up of devices (sdb, sdbf, etc..)
instead of paritions (sdb1, sdf1, etc...). And I don't know how to
mount a partitionless device. Do you?




==

# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5]
[raid4] [raid10]
md0 : active raid1 sdb[1] sde[6] sdh[5] sdg[4] sdf[3] sdl[2]
      976762496 blocks [7/6] [_UUUUUU]
      [=>...................]  resync =  8.3% (81881280/976762496)
finish=32031.5min speed=464K/sec

# /proc/sys/dev/raid/speed_limit_min
30000

# cat /proc/version
Linux version 2.6.24-8-server (buildd@yellow) (gcc version 4.2.3
(Ubuntu 4.2.3-1ubuntu2)) #1 SMP Thu Feb 14 20:42:20 UTC 2008

# mount | grep md0
/dev/md0 on /data1 type xfs (rw,noatime,logbufs=8,logbsize=256k)

# mdadm --query --detail /dev/md0
/dev/md0:
        Version : 00.90.03
  Creation Time : Thu Jan 31 18:56:43 2008
     Raid Level : raid1
     Array Size : 976762496 (931.51 GiB 1000.20 GB)
  Used Dev Size : 976762496 (931.51 GiB 1000.20 GB)
   Raid Devices : 7
  Total Devices : 6
Preferred Minor : 0
    Persistence : Superblock is persistent

    Update Time : Mon May 26 18:45:10 2008
          State : active, degraded, resyncing
 Active Devices : 6
Working Devices : 6
 Failed Devices : 0
  Spare Devices : 0

 Rebuild Status : 8% complete

           UUID : 9b7bc589:f34bfba0:72fed6b5:5ee48a3f
         Events : 0.3047399

    Number   Major   Minor   RaidDevice State
       0       0        0        0      removed
       1       8       16        1      active sync   /dev/sdb
       2       8      176        2      active sync   /dev/sdl
       3       8       80        3      active sync   /dev/sdf
       4       8       96        4      active sync   /dev/sdg
       5       8      112        5      active sync   /dev/sdh
       6       8       64        6      active sync   /dev/sde

Re: split off + mount a partitionless RAID1 device?

[Brad Campbell]

Jeff Breidenbach wrote:

> I'd love to split one of the drives off of the RAID1 and mount it
> directly. Then it can feed the ravenous Apache monster while the
> other five drives resync in peace. But foolishly and somewhat
> accidentally, this RAID is made up of devices (sdb, sdbf, etc..)
> instead of paritions (sdb1, sdf1, etc...). And I don't know how to
> mount a partitionless device. Do you?

Works just like mounting any other block device
You don't absolutely *require* a partition table on a device, you can get along fine without one if 
you are not worried about the kernel complaining from time to time about an invalid partition table.

mount /dev/sdb /mnt

Easy.

Brad
-- 
"Human beings, who are almost unique in having the ability
to learn from the experience of others, are also remarkable
for their apparent disinclination to do so." -- Douglas Adams

Re: split off + mount a partitionless RAID1 device?

[Jeff Breidenbach]

> mount /dev/sdb /mnt
>
> Easy.

mount -t xfs -o nouuid /dev/sdb /backup

Just for the record... xfs required a little bit of cajoling to avoid
complaints like these.

mount: unknown filesystem type 'linux_raid_member'
[85439.832068] XFS: Filesystem sdb has duplicate UUID - can't mount

Re: split off + mount a partitionless RAID1 device?

[NeilBrown]

On Tue, May 27, 2008 2:00 pm, Jeff Breidenbach wrote:
> Ok, let's say that a um... hypothetical friend of mine had a power
> outage this morning. This many disk RAID1 happily assembled and
> mounted (with only mild persuasion) but is now caught up in a
> monumentally slow resync. The resync is slow because /dev/md0 is
> getting hammered with read requests by a webserver. If I turn off
> Apache then resysnc speed shoots up to a respectable level, but I'd
> rather not do that.

Why not just leave the resync going in the background using whatever
bandwidth happens to be available?
It isn't as though your data is much at risk.

>
> I'd love to split one of the drives off of the RAID1 and mount it
> directly. Then it can feed the ravenous Apache monster while the
> other five drives resync in peace. But foolishly and somewhat
> accidentally, this RAID is made up of devices (sdb, sdbf, etc..)
> instead of paritions (sdb1, sdf1, etc...). And I don't know how to
> mount a partitionless device. Do you?

The traditional method is to use the "mount" command.
whole-devices are no different from partitions when it comes to
mounting them.

NeilBrown

Re: split off + mount a partitionless RAID1 device?

[Jeff Breidenbach]

> Why not just leave the resync going in the background using whatever
> bandwidth happens to be available?

Well, 32 thousand minutes is a long time. I figure sequestering
all those seeks from Apache would cut resync to a mere 8 hours.
Which would be nice because I'm trying to baby the array until
it is complete by holding off on writes.

However, for some unknown reason the array has recently sped up
a lot and is right now chugging along at speed=24485K/sec. So
I guess I'll just offer general thanks to the strange and mysterious
kernel gods and let things progress as is.

General thanks!

Re: split off + mount a partitionless RAID1 device?

[David Greaves]

Jeff Breidenbach wrote:
> However, for some unknown reason the array has recently sped up
> a lot and is right now chugging along at speed=24485K/sec. So
> I guess I'll just offer general thanks to the strange and mysterious
> kernel gods and let things progress as is.

Maybe most of what apache needs is now cached...

David