Raid over USB disks and eject

Raid over USB disks and eject

[Salatiel Filho]

Hi guys,  i have an external USB enclosure with 4 disks  in RAID5 mode
using mdadm. I dont know exactly why but if i use this disks for
massive I/O for lots of hours (over 12 hours), linux usually eject
some or all disks (causing a faulty in the raid) and remaps them: sdb
-> sdf , sdc -> sdg , sdd -> sdh , sde -> sdi
Well , this was never exactly a problem because most of the time those
disks are idle (just backup), but i had a real disk failure and i will
replace the disk. But probably the rebuild process will take a lot
more than 12 hours so the chances are that some disk will be ejected
and i will be in a trouble situation where a raid5 rebuilding process
will probably be stopped by an apparently failure. So my question is:

If i get into that situation , will i be able to disassemble and
reassemble the disk using the three disks i know are 'fine' and
restart the rebuild process from where it stopped ? I have internal
bitmap enabled.

Is it possible to somehow freeze the rebuild process ? So i could
pause the process for a few hours after 8 hours of work for example.

The raid will be assembled but umounted during the rebuild proccess.

I am using kernel 3.2.30.

Thanks !!!


[]'s
Salatiel

Re: Raid over USB disks and eject

[Benjamin ESTRABAUD]

Hi!

You could setup a cron job that will pause the resync by writing to sysfs:

If your array if "md0":

echo "idle" > /sys/block/md0/md/sync_action

However, this is probably not necessary as the resynced stripes will be 
marked in the bitmap, meaning that if your rebuild fails, re-assembling 
the array should restart the resync from where it stopped.

Regards,
Ben.

On 12/11/12 12:44, Salatiel Filho wrote:
> Hi guys,  i have an external USB enclosure with 4 disks  in RAID5 mode
> using mdadm. I dont know exactly why but if i use this disks for
> massive I/O for lots of hours (over 12 hours), linux usually eject
> some or all disks (causing a faulty in the raid) and remaps them: sdb
> -> sdf , sdc -> sdg , sdd -> sdh , sde -> sdi
> Well , this was never exactly a problem because most of the time those
> disks are idle (just backup), but i had a real disk failure and i will
> replace the disk. But probably the rebuild process will take a lot
> more than 12 hours so the chances are that some disk will be ejected
> and i will be in a trouble situation where a raid5 rebuilding process
> will probably be stopped by an apparently failure. So my question is:
>
> If i get into that situation , will i be able to disassemble and
> reassemble the disk using the three disks i know are 'fine' and
> restart the rebuild process from where it stopped ? I have internal
> bitmap enabled.
>
> Is it possible to somehow freeze the rebuild process ? So i could
> pause the process for a few hours after 8 hours of work for example.
>
> The raid will be assembled but umounted during the rebuild proccess.
>
> I am using kernel 3.2.30.
>
> Thanks !!!
>
>
> []'s
> Salatiel
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message tomajordomo@vger.kernel.org
> More majordomo info athttp://vger.kernel.org/majordomo-info.html
>

Re: Raid over USB disks and eject

[Salatiel Filho]

Hi Benjamin, thanks for the tip. I will try this later and post the
results here.

[]'s
Salatiel


On Mon, Nov 12, 2012 at 12:52 PM, Benjamin ESTRABAUD <be@mpstor.com> wrote:
> Hi!
>
> You could setup a cron job that will pause the resync by writing to sysfs:
>
> If your array if "md0":
>
> echo "idle" > /sys/block/md0/md/sync_action
>
> However, this is probably not necessary as the resynced stripes will be
> marked in the bitmap, meaning that if your rebuild fails, re-assembling the
> array should restart the resync from where it stopped.
>
> Regards,
> Ben.
>
>
> On 12/11/12 12:44, Salatiel Filho wrote:
>
> Hi guys,  i have an external USB enclosure with 4 disks  in RAID5 mode
> using mdadm. I dont know exactly why but if i use this disks for
> massive I/O for lots of hours (over 12 hours), linux usually eject
> some or all disks (causing a faulty in the raid) and remaps them: sdb
> -> sdf , sdc -> sdg , sdd -> sdh , sde -> sdi
> Well , this was never exactly a problem because most of the time those
> disks are idle (just backup), but i had a real disk failure and i will
> replace the disk. But probably the rebuild process will take a lot
> more than 12 hours so the chances are that some disk will be ejected
> and i will be in a trouble situation where a raid5 rebuilding process
> will probably be stopped by an apparently failure. So my question is:
>
> If i get into that situation , will i be able to disassemble and
> reassemble the disk using the three disks i know are 'fine' and
> restart the rebuild process from where it stopped ? I have internal
> bitmap enabled.
>
> Is it possible to somehow freeze the rebuild process ? So i could
> pause the process for a few hours after 8 hours of work for example.
>
> The raid will be assembled but umounted during the rebuild proccess.
>
> I am using kernel 3.2.30.
>
> Thanks !!!
>
>
> []'s
> Salatiel
> --
> 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: Raid over USB disks and eject

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 2460 bytes --]

On Mon, 12 Nov 2012 16:07:28 +0000 Benjamin ESTRABAUD <be@mpstor.com> wrote:

> Hi!
> 
> You could setup a cron job that will pause the resync by writing to sysfs:
> 
> If your array if "md0":
> 
> echo "idle" > /sys/block/md0/md/sync_action

"idle" is not sufficient.  It will stop the recovery, but then it will
immediately start again.
You need "frozen".

NeilBrown


> 
> However, this is probably not necessary as the resynced stripes will be 
> marked in the bitmap, meaning that if your rebuild fails, re-assembling 
> the array should restart the resync from where it stopped.
> 
> Regards,
> Ben.
> 
> On 12/11/12 12:44, Salatiel Filho wrote:
> > Hi guys,  i have an external USB enclosure with 4 disks  in RAID5 mode
> > using mdadm. I dont know exactly why but if i use this disks for
> > massive I/O for lots of hours (over 12 hours), linux usually eject
> > some or all disks (causing a faulty in the raid) and remaps them: sdb
> > -> sdf , sdc -> sdg , sdd -> sdh , sde -> sdi
> > Well , this was never exactly a problem because most of the time those
> > disks are idle (just backup), but i had a real disk failure and i will
> > replace the disk. But probably the rebuild process will take a lot
> > more than 12 hours so the chances are that some disk will be ejected
> > and i will be in a trouble situation where a raid5 rebuilding process
> > will probably be stopped by an apparently failure. So my question is:
> >
> > If i get into that situation , will i be able to disassemble and
> > reassemble the disk using the three disks i know are 'fine' and
> > restart the rebuild process from where it stopped ? I have internal
> > bitmap enabled.
> >
> > Is it possible to somehow freeze the rebuild process ? So i could
> > pause the process for a few hours after 8 hours of work for example.
> >
> > The raid will be assembled but umounted during the rebuild proccess.
> >
> > I am using kernel 3.2.30.
> >
> > Thanks !!!
> >
> >
> > []'s
> > Salatiel
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> > the body of a message tomajordomo@vger.kernel.org
> > More majordomo info athttp://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


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: Raid over USB disks and eject

[Salatiel Filho]

Hi NeilBrown, the frozen option did the trick. Thanks ! So far rebuild
with no problems. 2 hours rebuilding , 15min sleeping in a loop ...

[]'s
Salatiel


On Mon, Nov 12, 2012 at 5:28 PM, NeilBrown <neilb@suse.de> wrote:
> On Mon, 12 Nov 2012 16:07:28 +0000 Benjamin ESTRABAUD <be@mpstor.com> wrote:
>
>> Hi!
>>
>> You could setup a cron job that will pause the resync by writing to sysfs:
>>
>> If your array if "md0":
>>
>> echo "idle" > /sys/block/md0/md/sync_action
>
> "idle" is not sufficient.  It will stop the recovery, but then it will
> immediately start again.
> You need "frozen".
>
> NeilBrown
>
>
>>
>> However, this is probably not necessary as the resynced stripes will be
>> marked in the bitmap, meaning that if your rebuild fails, re-assembling
>> the array should restart the resync from where it stopped.
>>
>> Regards,
>> Ben.
>>
>> On 12/11/12 12:44, Salatiel Filho wrote:
>> > Hi guys,  i have an external USB enclosure with 4 disks  in RAID5 mode
>> > using mdadm. I dont know exactly why but if i use this disks for
>> > massive I/O for lots of hours (over 12 hours), linux usually eject
>> > some or all disks (causing a faulty in the raid) and remaps them: sdb
>> > -> sdf , sdc -> sdg , sdd -> sdh , sde -> sdi
>> > Well , this was never exactly a problem because most of the time those
>> > disks are idle (just backup), but i had a real disk failure and i will
>> > replace the disk. But probably the rebuild process will take a lot
>> > more than 12 hours so the chances are that some disk will be ejected
>> > and i will be in a trouble situation where a raid5 rebuilding process
>> > will probably be stopped by an apparently failure. So my question is:
>> >
>> > If i get into that situation , will i be able to disassemble and
>> > reassemble the disk using the three disks i know are 'fine' and
>> > restart the rebuild process from where it stopped ? I have internal
>> > bitmap enabled.
>> >
>> > Is it possible to somehow freeze the rebuild process ? So i could
>> > pause the process for a few hours after 8 hours of work for example.
>> >
>> > The raid will be assembled but umounted during the rebuild proccess.
>> >
>> > I am using kernel 3.2.30.
>> >
>> > Thanks !!!
>> >
>> >
>> > []'s
>> > Salatiel
>> > --
>> > To unsubscribe from this list: send the line "unsubscribe linux-raid" in
>> > the body of a message tomajordomo@vger.kernel.org
>> > More majordomo info athttp://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
>