Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.os.linux.misc > #1400

Re: RAID: prevent out-of-sync with unstable connections

Show all headers | View raw

David Brown <david.brown@removethis.hesbynett.no> wrote:
> On 12/06/11 13:30, Lasse Kliemann wrote:
>> Is there a way to configure a RAID, say level 1 or 5, so that
>> whenever one drive fails, nothing is done to the remaining drives
>> that would bring then out of sync relative to the failed drive?
>>
>> If the failure was just due to a shaky cable connection, then the
>> array could be re-started after fixing the cables without
>> requiring a re-sync.
>>
>> There is the --no-degraded option to mdadm which is helpful when
>> the cable problem is already present when the array is assembled.
>> But it does not help if something happens during operation.
>>
>> I have a portable RAID that is connected via USB or eSATA to
>> different machines. On some machines, USB is unstable at times,
>> leading to the array becoming degraded. The following re-sync is
>> very time-consuming.
> 
> I think the only thing you could do would be to put the array in 
> read-only mode (certainly you can't write to the array without it 
> becoming out-of-sync).  However, if you were writing to the array when 
> it failed, the writes will be completed to the array before you get a 
> chance to make it read-only.
> 
> In other words, no, there is no way to stop them getting out of sync.
> 
> However, I think you could come a long way with a write intent bitmap. 
> I'm not sure of the details here, but I believe that will let the array 
> track changes to disks after the disconnection.  You can then re-connect 
> and only re-sync the changed data.
> 
> But you'll have to do a bit of reading to see exactly how to achieve this.
 
A write intent bitmap was what I was looking for. Thank you!

It is explained here:
  http://etbe.coker.com.au/2008/01/28/write-intent-bitmaps/

There can be a performance loss involved. In my case, I found 
that out to be minor with a couple of quick benchmarks. It is 
also influenced by the chunk size used for the bitmap, which I 
set to 1GiB.

I was also daring enough to risk another 15 hours of resync by 
marking a component faulty and re-adding it. Resync finished 
within a couple of minutes. That's great!

Back to comp.os.linux.misc | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

RAID: prevent out-of-sync with unstable connections Lasse Kliemann <lasse-usenet-2011@mail.plastictree.net> - 2011-06-12 13:30 +0200
  Re: RAID: prevent out-of-sync with unstable connections David Brown <david.brown@removethis.hesbynett.no> - 2011-06-12 15:33 +0200
    Re: RAID: prevent out-of-sync with unstable connections Lasse Kliemann <lasse-usenet-2011@mail.plastictree.net> - 2011-06-12 17:21 +0200
      Re: RAID: prevent out-of-sync with unstable connections David Brown <david.brown@removethis.hesbynett.no> - 2011-06-12 20:13 +0200
        Re: Disk access every few seconds (was: RAID: prevent out-of-sync with unstable connections) Lasse Kliemann <lasse-usenet-2011@mail.plastictree.net> - 2011-06-14 09:05 +0200
          Re: Disk access every few seconds David Brown <david@westcontrol.removethisbit.com> - 2011-06-14 13:39 +0200
            Re: Disk access every few seconds Lasse Kliemann <lasse-usenet-2011@mail.plastictree.net> - 2011-06-15 08:51 +0200
              Re: Disk access every few seconds David Brown <david@westcontrol.removethisbit.com> - 2011-06-15 11:25 +0200
                Re: Disk access every few seconds "Peter J. Holzer" <hjp-usenet2@hjp.at> - 2011-06-17 15:34 +0200
                Re: [SOLVED] Disk access every few seconds (was: Disk access every few seconds) Lasse Kliemann <lasse-usenet-2011@mail.plastictree.net> - 2011-06-26 19:11 +0200
                Re: [SOLVED] Disk access every few seconds Lasse Kliemann <lasse-usenet-2011@mail.plastictree.net> - 2011-06-27 22:24 +0200

csiph-web