Some raid1 questions

Some raid1 questions

[Yves Glodt]

Hello,

I have my home partition on a raid1 device, and there is an ext3 
filesystem on it. Some more information:

fstab:
----------------
/dev/hdc1 / ext3 defaults,noatime,nodiratime
/dev/md0 /home ext3 defaults,noatime,nodiratime,data=writeback

raidtab:
---------------
raiddev                 /dev/md0
raid-level              1
persistent-superblock   1
nr-raid-disks           2
nr-spare-disks          0
device                  /dev/hda2
raid-disk               0
device                  /dev/hdc2
raid-disk               1


Recently I had some freezes (due to nvidia...?), and everytime I had to 
make a hard reset.

Everytime then the raid resynced completely which took about 1.5 hours.
One time I lost some kde settings when this happened.
The filesystem did not complain about an unclean shutdown, it seems it 
did not see on bootup that the last shutdown was unclean.

Now to my questions:
Is this the normal behaviour?
Shouldn't the filesystem take care of that?
What is the point of having a journaling filesystem on a mirror?
Is there anything I can change to have the filesystem handle unclean 
shutdowns?

TIA,
Yves


-- 
Linux 2.4.24 #5 Mon Jan 5 19:38:06 CET 2004 i686
 14:25:55 up  4:14,  1 user,  load average: 0.43, 0.40, 0.32

Re: Some raid1 questions

[Neil Brown]

On Sunday February 1, yg@mind.lu wrote:
> Hello,
> 
> I have my home partition on a raid1 device, and there is an ext3 
> filesystem on it. Some more information:
...
> 
> Recently I had some freezes (due to nvidia...?), and everytime I had to 
> make a hard reset.
> 
> Everytime then the raid resynced completely which took about 1.5 hours.
> One time I lost some kde settings when this happened.
> The filesystem did not complain about an unclean shutdown, it seems it 
> did not see on bootup that the last shutdown was unclean.
> 
> Now to my questions:
> Is this the normal behaviour?
Which behaviour?

  1.5 hour resync -- maybe. It depends on the speed and size of the
    drives.
  lost kde settings -- maybe. It depends on how well kde copes with
    sudden system failure.
  The filesystem didn't complain -- Yes.  ext3 doesn't complain about
    an unclean shutdown.  It just replays the journal and keeps
    working. 

> Shouldn't the filesystem take care of that?

  Take care of what?  I suspect it did take care of everything it
  could. 

> What is the point of having a journaling filesystem on a mirror?
  Lots.  A journaling filesystem guards against unclean shutdown.  A
  mirror guards against drive failure.  These are compliementary.

> Is there anything I can change to have the filesystem handle unclean 
> shutdowns?

 I think it did handle the unclean shutdown just fine.
 The loss of kde settings is probably dud to KDE note being careful
 enough when updating it's settings file.

NeilBrown

> 
> TIA,
> Yves
> 
> 
> -- 
> Linux 2.4.24 #5 Mon Jan 5 19:38:06 CET 2004 i686
>  14:25:55 up  4:14,  1 user,  load average: 0.43, 0.40, 0.32
> 
> -
> 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

Re: Some raid1 questions

[Yves Glodt]

> On Sunday February 1, yg@mind.lu wrote:
>> Hello,
>>
>> I have my home partition on a raid1 device, and there is an ext3
>> filesystem on it. Some more information:
> ...
>>
>> Recently I had some freezes (due to nvidia...?), and everytime I had
>> to  make a hard reset.
>>
>> Everytime then the raid resynced completely which took about 1.5
>> hours. One time I lost some kde settings when this happened.

>> The filesystem did not complain about an unclean shutdown, it seems it
>>  did not see on bootup that the last shutdown was unclean.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

>> Now to my questions:
>> Is this the normal behaviour?
> Which behaviour?

Sorry for being unclear...
After an unclean shutdown, my raid always rebuilds, and the filesystem never
replays the journal. I just does not see that it was not properly unmounted.

Now again :)... is that the normal behaviour?


>   1.5 hour resync -- maybe. It depends on the speed and size of the
>     drives.

ok

>   lost kde settings -- maybe. It depends on how well kde copes with
>     sudden system failure.

ok

>   The filesystem didn't complain -- Yes.  ext3 doesn't complain about
>     an unclean shutdown.  It just replays the journal and keeps
>     working.

It did not replay anything in these cases.

>> Shouldn't the filesystem take care of that?
>
>   Take care of what?  I suspect it did take care of everything it
>   could.
>
>> What is the point of having a journaling filesystem on a mirror?
>   Lots.  A journaling filesystem guards against unclean shutdown.  A
> mirror guards against drive failure.  These are compliementary.
>
>> Is there anything I can change to have the filesystem handle unclean
>> shutdowns?
>
>  I think it did handle the unclean shutdown just fine.
>  The loss of kde settings is probably dud to KDE note being careful
> enough when updating it's settings file.
>
> NeilBrown
>
>>
>> TIA,
>> Yves
>>
>>
>> --
>> Linux 2.4.24 #5 Mon Jan 5 19:38:06 CET 2004 i686
>>  14:25:55 up  4:14,  1 user,  load average: 0.43, 0.40, 0.32
>>
>> -
>> 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

Re: Some raid1 questions

[Neil Brown]

On Monday February 2, yg@mind.lu wrote:
> 
> Sorry for being unclear...
> After an unclean shutdown, my raid always rebuilds, and the filesystem never
> replays the journal. I just does not see that it was not properly
> unmounted.

That is certainly odd.  Are you sure that is what is happening?
There are two ways that the journal can be replayed:
   1/ fsck can replay it or
   2/ ext3 in the kernel can replay it.

The former will not appear in any logs and will only be seen if you
watch the boot-time messages go past.  The latter will cause a kernel
message to be logged.

Were you watching for fsck messages, or only looking in logs?

NeilBrown

Re: Some raid1 questions

[Yves Glodt]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Monday 02 February 2004 22:48, Neil Brown wrote:
> On Monday February 2, yg@mind.lu wrote:
> > Sorry for being unclear...
> > After an unclean shutdown, my raid always rebuilds, and the
> > filesystem never replays the journal. I just does not see that it
> > was not properly unmounted.
>
> That is certainly odd.  Are you sure that is what is happening?
> There are two ways that the journal can be replayed:
>    1/ fsck can replay it or
>    2/ ext3 in the kernel can replay it.
>
> The former will not appear in any logs and will only be seen if you
> watch the boot-time messages go past.  The latter will cause a kernel
> message to be logged.
>
> Were you watching for fsck messages, or only looking in logs?

I was watching the boot messages. I use the "quiet" kernel parameter, so 
the first thing I really saw was that the raid complained about not 
being clean.
But, now I checked my /var/log/messages, and I saw that in about 4 of 5 
times when I had a freeze, on the next boot, the ext3 driver seemed to 
recover something, look at this:

Jan 27 17:45:40 Valhalla kernel: md: autorun ...
Jan 27 17:45:40 Valhalla kernel: md: considering hdc2 ...
Jan 27 17:45:40 Valhalla kernel: md:  adding hdc2 ...
Jan 27 17:45:40 Valhalla kernel: md:  adding hda2 ...
Jan 27 17:45:40 Valhalla kernel: md: created md0
Jan 27 17:45:40 Valhalla kernel: md: bind<hda2,1>
Jan 27 17:45:40 Valhalla kernel: md: bind<hdc2,2>
Jan 27 17:45:40 Valhalla kernel: md: running: <hdc2><hda2>
Jan 27 17:45:40 Valhalla kernel: md: hdc2's event counter: 00001115
Jan 27 17:45:40 Valhalla kernel: md: hda2's event counter: 00001115
Jan 27 17:45:40 Valhalla kernel: md: RAID level 1 does not need 
chunksize! Continuing anyway.
Jan 27 17:45:40 Valhalla kernel: md: raid1 personality registered as nr 
3
Jan 27 17:45:40 Valhalla kernel: md0: max total readahead window set to 
124k
Jan 27 17:45:40 Valhalla kernel: md0: 1 data-disks, max readahead per 
data-disk: 124k
Jan 27 17:45:40 Valhalla kernel: raid1: device hdc2 operational as 
mirror 1
Jan 27 17:45:40 Valhalla kernel: raid1: device hda2 operational as 
mirror 0
Jan 27 17:45:40 Valhalla kernel: raid1: raid set md0 not clean; 
reconstructing mirrors
Jan 27 17:45:40 Valhalla kernel: raid1: raid set md0 active with 2 out 
of 2 mirrors
Jan 27 17:45:40 Valhalla kernel: md: updating md0 RAID superblock on 
device
Jan 27 17:45:40 Valhalla kernel: md: hdc2 [events: 00001116]<6>(write) 
hdc2's sb offset: 29302464
Jan 27 17:45:40 Valhalla kernel: md: syncing RAID array md0
Jan 27 17:45:40 Valhalla kernel: md: minimum _guaranteed_ reconstruction 
speed: 100 KB/sec/disc.
Jan 27 17:45:40 Valhalla kernel: md: using maximum available idle IO 
bandwith (but not more than 100000 KB/sec) for reconstruction.
Jan 27 17:45:40 Valhalla kernel: md: using 124k window, over a total of 
29302464 blocks.
Jan 27 17:45:40 Valhalla kernel: md: hda2 [events: 00001116]<6>(write) 
hda2's sb offset: 29302464
Jan 27 17:45:40 Valhalla kernel: md: ... autorun DONE.
Jan 27 17:45:40 Valhalla kernel: kjournald starting.  Commit interval 5 
seconds
Jan 27 17:45:40 Valhalla kernel: EXT3 FS 2.4-0.9.19, 19 August 2002 on 
md(9,0), internal journal
Jan 27 17:45:40 Valhalla kernel: EXT3-fs: recovery complete.
Jan 27 17:45:40 Valhalla kernel: EXT3-fs: mounted filesystem with 
writeback data mode.







So it seems that the filesystem is recovering (in 4 of 5 times):
Here the example where it did not:

Jan 19 18:57:49 Valhalla kernel: md: autorun ...
Jan 19 18:57:49 Valhalla kernel: md: considering hdc2 ...
Jan 19 18:57:49 Valhalla kernel: md:  adding hdc2 ...
Jan 19 18:57:49 Valhalla kernel: md:  adding hda2 ...
Jan 19 18:57:49 Valhalla kernel: md: created md0
Jan 19 18:57:49 Valhalla kernel: md: bind<hda2,1>
Jan 19 18:57:49 Valhalla kernel: md: bind<hdc2,2>
Jan 19 18:57:49 Valhalla kernel: md: running: <hdc2><hda2>
Jan 19 18:57:49 Valhalla kernel: md: hdc2's event counter: 000010f8
Jan 19 18:57:49 Valhalla kernel: md: hda2's event counter: 000010f8
Jan 19 18:57:49 Valhalla kernel: md: RAID level 1 does not need 
chunksize! Continuing anyway.
Jan 19 18:57:49 Valhalla kernel: md: raid1 personality registered as nr 
3
Jan 19 18:57:49 Valhalla kernel: md0: max total readahead window set to 
124k
Jan 19 18:57:49 Valhalla kernel: md0: 1 data-disks, max readahead per 
data-disk: 124k
Jan 19 18:57:49 Valhalla kernel: raid1: device hdc2 operational as 
mirror 1
Jan 19 18:57:49 Valhalla kernel: raid1: device hda2 operational as 
mirror 0
Jan 19 18:57:49 Valhalla kernel: raid1: raid set md0 not clean; 
reconstructing mirrors
Jan 19 18:57:49 Valhalla kernel: raid1: raid set md0 active with 2 out 
of 2 mirrors
Jan 19 18:57:49 Valhalla kernel: md: updating md0 RAID superblock on 
device
Jan 19 18:57:49 Valhalla kernel: md: hdc2 [events: 000010f9]<6>(write) 
hdc2's sb offset: 29302464
Jan 19 18:57:49 Valhalla kernel: md: syncing RAID array md0
Jan 19 18:57:49 Valhalla kernel: md: minimum _guaranteed_ reconstruction 
speed: 100 KB/sec/disc.
Jan 19 18:57:49 Valhalla kernel: md: using maximum available idle IO 
bandwith (but not more than 100000 KB/sec) for reconstruction.
Jan 19 18:57:49 Valhalla kernel: md: using 124k window, over a total of 
29302464 blocks.
Jan 19 18:57:49 Valhalla kernel: md: hda2 [events: 000010f9]<6>(write) 
hda2's sb offset: 29302464
Jan 19 18:57:49 Valhalla kernel: md: ... autorun DONE.
Jan 19 18:57:49 Valhalla kernel: kjournald starting.  Commit interval 5 
seconds
Jan 19 18:57:49 Valhalla kernel: EXT3 FS 2.4-0.9.19, 19 August 2002 on 
md(9,0), internal journal
Jan 19 18:57:49 Valhalla kernel: EXT3-fs: mounted filesystem with 
writeback data mode.






Ok, I guess the filesystem just behaves fine, well but still I wonder if 
it's "normal" that the raid reconstructs also after *every* unclean 
shutdown.
thank you for your support,
Yves


> 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

- -- 
Linux 2.4.24 #5 Mon Jan 5 19:38:06 CET 2004 i686
 23:03:40 up  4:54,  1 user,  load average: 0.02, 0.08, 0.15
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFAHs78fmxiTOp0sQYRAgc4AJ4orJbkdYopVysZo7nq1f9hGhz9bgCeKjQX
or6pSSaGIeJFsTAqqw1Gg3o=
=6287
-----END PGP SIGNATURE-----