Well I browsed through the sources of latest mdadm version at night
instead of sleeping :) and was searching how it got clean flag set to
0. And was not sure exactly about that line there and from which
device it gets clean state. So the bug was that it can get it from not
current device? It makes sense because first device identified by
mdadm is old md101 device.

So will it work then if I use 3.3 and somehow dont give it md101
device? By stopping it before -A call or by manually specifying other
drives? Or you really recommend to build latest version of mdadm?

What is expected behaviour with 3.3.1+? Can it be started with all
devices and it should automatically start to recover md101? If so what
is best way, to start it without md101 and then use -re-add to add it,
start it without md101 and use -add or start it also with md101?

Thank you very much.

2014-09-22 8:56 GMT+02:00 NeilBrown [off-list ref]:

On Mon, 22 Sep 2014 08:34:21 +0200 Patrik Horník [off-list ref] wrote:

quoted

- Well what is exact meaning of --no-degraded then? Because I am using
it also on RAID6 arrays that are missing one drive and mdadm starts
them. I thought until today that it is against assembling for example
RAID6 array with missing more than two drives or to be more precise
array with number of drives it used last time. (I did not look at the
code what does it exactly. It is mdadm 3.3 on Debian.)

Sorry, I confused myself.
"--no-degraded" means "Only start the array if all expected devices are
present".
So if the array "knows" that one device is missing, it will start if all
other devices are present.  But if it "thinks" that all devices are working,
then it will only start if all the devices ar there.

quoted

- Well array was shutdown cleanly manually by mdadm -S. Cant the not
clean classification be result of md101 device between find devices or
result of first two assemble tries?

If the state still says "Clean" (which it does, thanks), the mdadm should
treat it as 'clean'.

I think you are probably hitting the bug fixed by

 http://git.neil.brown.name/?p=mdadm.git;a=commitdiff;h=56bbc588f7f0f3bdd3ec23f02109b427c1d3b8f1

which is in 3.3.1.

So a new version of mdadm should fix it.

NeilBrown

quoted

- Anyway as I mentioned superblock on all five devices has clean state. Example:
/dev/sdk1:
          Magic : XXXXXXX
        Version : 1.2
    Feature Map : 0x1
     Array UUID : XXXXXXXXXXXXXXXXXXXXX
           Name :
  Creation Time : Thu Aug XXXXXXXX
     Raid Level : raid6
   Raid Devices : 6

 Avail Dev Size : 5860268943 (2794.39 GiB 3000.46 GB)
     Array Size : 11720536064 (11177.57 GiB 12001.83 GB)
  Used Dev Size : 5860268032 (2794.39 GiB 3000.46 GB)
    Data Offset : 262144 sectors
   Super Offset : 8 sectors
   Unused Space : before=262056 sectors, after=911 sectors
          State : clean
    Device UUID : YYYYYYYYYYYYYYYYYYYY

Internal Bitmap : 8 sectors from superblock
    Update Time : Mon Sep 22 02:23:45 2014
  Bad Block Log : 512 entries available at offset 72 sectors
       Checksum : ZZZZZZZZ - correct
         Events : EEEEEE

         Layout : left-symmetric
     Chunk Size : 512K

   Device Role : Active device 4
   Array State : AAAAA. ('A' == active, '.' == missing, 'R' == replacing)

- md101 has Events count lower by 16 than others devices.

- Please I need little more assurance what is exact state of array and
explain why it is behaving as it is behaving, so I can be sure what
steps are needed and what happens. The data on array is important.
Patrik Horník
šéfredaktor www.DSL.sk
Tel.: +421 905 385 666
Email: patrik@dsl.sk


2014-09-22 5:19 GMT+02:00 NeilBrown [off-list ref]:

quoted

On Mon, 22 Sep 2014 04:11:20 +0200 Patrik Horník [off-list ref] wrote:

quoted

Hello Neil,

I've got this situation unfamiliar to me on RAID6 array md1 with important data.

- It is RAID6 with 6 devices, 5 are partitions and 1 is another RAID0
array md101 from two smaller drives. One of the smaller drives froze,
so md101 got kicked out from md1 and marked as faulty in md1. After
while I've stopped md1 without removing md101 from it first. Then I
rebooted and assembled md101.

- First I tried mdadm -A --no-degraded -u UUID /dev/md1 but got
"mdadm: /dev/md1 assembled from 5 drives (out of 6), but not started."
so I stopped the md1.

- Second time I started it with -v and got:

mdadm: /dev/md101 is identified as a member of /dev/md1, slot 5.
mdadm: /dev/sdk1 is identified as a member of /dev/md1, slot 4.
mdadm: /dev/sdi1 is identified as a member of /dev/md1, slot 1.
mdadm: /dev/sdh1 is identified as a member of /dev/md1, slot 2.
mdadm: /dev/sdg1 is identified as a member of /dev/md1, slot 0.
mdadm: /dev/sde1 is identified as a member of /dev/md1, slot 3.
mdadm: added /dev/sdi1 to /dev/md1 as 1
mdadm: added /dev/sdh1 to /dev/md1 as 2
mdadm: added /dev/sde1 to /dev/md1 as 3
mdadm: added /dev/sdk1 to /dev/md1 as 4
mdadm: added /dev/md101 to /dev/md1 as 5 (possibly out of date)
mdadm: added /dev/sdg1 to /dev/md1 as 0
mdadm: /dev/md1 assembled from 5 drives (out of 6), but not started.

- On third time I tried without --nodegraded with mdadm -A -v -u UUID
/dev/md1. This is what I've got:

mdadm: /dev/md101 is identified as a member of /dev/md1, slot 5.
mdadm: /dev/sdk1 is identified as a member of /dev/md1, slot 4.
mdadm: /dev/sdi1 is identified as a member of /dev/md1, slot 1.
mdadm: /dev/sdh1 is identified as a member of /dev/md1, slot 2.
mdadm: /dev/sdg1 is identified as a member of /dev/md1, slot 0.
mdadm: /dev/sde1 is identified as a member of /dev/md1, slot 3.
mdadm: added /dev/sdi1 to /dev/md1 as 1
mdadm: added /dev/sdh1 to /dev/md1 as 2
mdadm: added /dev/sde1 to /dev/md1 as 3
mdadm: added /dev/sdk1 to /dev/md1 as 4
mdadm: added /dev/md101 to /dev/md1 as 5 (possibly out of date)
mdadm: added /dev/sdg1 to /dev/md1 as 0
mdadm: /dev/md1 assembled from 5 drives - not enough to start the
array while not clean - consider --force.

Array md1 has bitmap. All drive devices have all same Events, their
state is clean and Device Role is Active device. md101 has active
state and lower Events.

Is this expected behavior? My theory is that it is caused by md101 and
I should start array md1 without it (by for example stopping md101)
and then re-add it. Is that a case or is it something else?

Thanks.

Best regards,

Patrik

The array is clearly degraded as one of the devices failed and hasn't been
recovered yet, so using --nodegraded is counter productive, as you
discovered.

It appears that the array is also marked as 'dirty'.  That suggests that it
wasn't shut down cleanly.
What does "mdadm --examine" of some device show?

You probably need to re-assemble the array with --force like it suggests,
then add the failed device and let it recover.

NeilBrown

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html