[Gluster-users] Self-heal failed on new bricks. Or did it?

[Joe Julian]

On 03/12/2014 02:38 PM, John Gardeniers wrote:
> Hi All,
>
> Having received no replies overnight I thought to see what would happen
> if I reboot each of the replica servers (one at a time of course). The
> problem disappeared. Theories anyone?
>
> John
>
>
> On 12/03/14 15:46, John Gardeniers wrote:
>> Hi All,
>>
>> I'm new to Gluster and am therefore at the "suck it and see" stage,
>> trying different things to see it in operation. I have a replica pair
>> with a single volume of one brick on each server acting as a back end
>> for an RHEV test install.
>>
>> Today I added an additional brick on each of the replica pair with
>> "gluster volume add-brick testvol 192.168.19.20:/brick2
>> 192.168.19.21:/brick2 force", the force being necessary as brick2 is on
>> the system drive (remember, I'm just experimenting for now).
>>
>> While there are no other signs of errors, when I run "gluster volume
>> heal testvol info heal-failed" I get the following:
>>
>> Gathering Heal info on volume testvol has been successful
>>
>> Brick 192.168.19.20:/storage
>> Number of entries: 0
>>
>> Brick 192.168.19.21:/storage
>> Number of entries: 0
>>
>> Brick 192.168.19.20:/brick2
>> Number of entries: 4
>> at                    path on brick
>> -----------------------------------
>> 2014-03-12 03:59:50 /f67693a9-e255-4466-a5a9-298e54f0f7c7/dom_md
>> 2014-03-12 03:59:50 /f67693a9-e255-4466-a5a9-298e54f0f7c7/master/tasks
>> 2014-03-12 03:59:50 /f67693a9-e255-4466-a5a9-298e54f0f7c7/master/vms
>> 2014-03-12 03:59:50 /f67693a9-e255-4466-a5a9-298e54f0f7c7/master
>>
>> Brick 192.168.19.21:/brick2
>> Number of entries: 0
>>
>> I've rsynced brick2 from each gluster server to another box and diffed
>> them. There are absolutely no differences. So how come Gluster is saying
>> that the heal failed? More importantly, what can I do about it?
>>
>> Thanks,
>> John
>>
>> _______
You'll need to look into your log files to find more information. If a 
heal fails, the log file where that heal was attempted should show why 
(probably one of the glustershd.log files, but it could be a client 
mount log as well).