Feature request: ability to exchange drive without failing the old one

Feature request: ability to exchange drive without failing the old one

[Christian Pernegger]

Hi list!

Sometimes I want to exchange a drive in an array even if it hasn't
actually failed, e. g.
- in preperation for a grow operation: exchange-resync-exchange-resync-...-grow
- when removing old drives as part of regular maintenance
- ...

Currently one has to fail an active drive to trigger a resync to the
replacement (assuming it was added as a spare beforehand) and thus
suffer the loss of fault tolerance and performance until the sync is
complete.

Suggestion: allow adding drives as an explicit replacement for an
active one. The active drive would then be copied to the new one in
the background, effectively replacing the component device with a
mirror. Bonus points if the resulting setup is a valid mode of
operation and not just a temporary hack. I suppose one could see the
feature as a special case of the reshaping that's all the rage now :)

Thanks for listening,

Christian

Re: Feature request: ability to exchange drive without failing the old one

[Bill Davidsen]

Christian Pernegger wrote:
> Hi list!
>
> Sometimes I want to exchange a drive in an array even if it hasn't
> actually failed, e. g.
> - in preperation for a grow operation: exchange-resync-exchange-resync-...-grow
> - when removing old drives as part of regular maintenance
> - ...
>
> Currently one has to fail an active drive to trigger a resync to the
> replacement (assuming it was added as a spare beforehand) and thus
> suffer the loss of fault tolerance and performance until the sync is
> complete.
>
> Suggestion: allow adding drives as an explicit replacement for an
> active one. The active drive would then be copied to the new one in
> the background, effectively replacing the component device with a
> mirror. Bonus points if the resulting setup is a valid mode of
> operation and not just a temporary hack. I suppose one could see the
> feature as a special case of the reshaping that's all the rage now :)
>   

I think what you want is a subset of the problem many of us face, moving 
a member of an array to a larger partition on a (much) larger drive. 
That migration, followed by a grow, is getting more appealing, as 
various drive drop to "prize in a cereal box" pricing. And as data 
grows, backup to something other than disk is getting pretty 
impractical, resulting in unit upgrades to save space, time, power, etc, 
etc.

I agree with the need, I have several ugly methods of doing this which 
boil down to replacing a member with a raid-1 of the member and the new 
drive. I have done it several ways, including the "I can't fit another 
drive in the box" solution, using nbd and write mostly. Uglier than a 
hedgehog's ass, but it worked... eventually.

-- 
Bill Davidsen <davidsen@tmr.com>
  Unintended results are the well-earned reward for incompetence.

Re: Feature request: ability to exchange drive without failing the old one

[Majed B.]

Isn't this called an active hot spare?

I think some hardware RAID controllers offer this feature: Keeping the
spare in sync with the whole array, so when a failure happens, the
spare takes its place immediately.

That'd be an awesome feature to have!

On Tue, Oct 13, 2009 at 11:38 PM, Bill Davidsen <davidsen@tmr.com> wrote:
> Christian Pernegger wrote:
>>
>> Hi list!
>>
>> Sometimes I want to exchange a drive in an array even if it hasn't
>> actually failed, e. g.
>> - in preperation for a grow operation:
>> exchange-resync-exchange-resync-...-grow
>> - when removing old drives as part of regular maintenance
>> - ...
>>
>> Currently one has to fail an active drive to trigger a resync to the
>> replacement (assuming it was added as a spare beforehand) and thus
>> suffer the loss of fault tolerance and performance until the sync is
>> complete.
>>
>> Suggestion: allow adding drives as an explicit replacement for an
>> active one. The active drive would then be copied to the new one in
>> the background, effectively replacing the component device with a
>> mirror. Bonus points if the resulting setup is a valid mode of
>> operation and not just a temporary hack. I suppose one could see the
>> feature as a special case of the reshaping that's all the rage now :)
>>
>
> I think what you want is a subset of the problem many of us face, moving a
> member of an array to a larger partition on a (much) larger drive. That
> migration, followed by a grow, is getting more appealing, as various drive
> drop to "prize in a cereal box" pricing. And as data grows, backup to
> something other than disk is getting pretty impractical, resulting in unit
> upgrades to save space, time, power, etc, etc.
>
> I agree with the need, I have several ugly methods of doing this which boil
> down to replacing a member with a raid-1 of the member and the new drive. I
> have done it several ways, including the "I can't fit another drive in the
> box" solution, using nbd and write mostly. Uglier than a hedgehog's ass, but
> it worked... eventually.
>
> --
> Bill Davidsen <davidsen@tmr.com>
>  Unintended results are the well-earned reward for incompetence.
>
> --
> 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
>



-- 
       Majed B.
--
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