[sheepdog-users] failed to read inode header

[Maxim Terletskiy]

Hello

Yes, it triggered recovery, but after recovery I still see "no majority" 
messages.

Before reweight I checked couple of times my test image and check 
reported me about three "no majority" objects. After reweight I see two 
messages. Searched with find to determine which node stores this pieces 
and see that one node have two copies of this pieces on different disks. 
Is it normal? Maybe that is why recovery don't help with "no majority"?

31.05.2014 18:47, Liu Yuan пишет:
> On Sat, May 31, 2014 at 06:06:21PM +0400, Maxim Terletskiy wrote:
>> Tried to start "dog cluster recover force":
>> failed to execute request, Cluster is running|halted and cannot be
>> force recovered
>>
>> Well, cluster is running, vms is running. I need to shutdown all
>> vms? Or I need to to "dog cluster shutdown"?
> Hmm, you might try
>
> $ dog cluster reweight # this also trigger a cluster-wide recovery.
>
> Thanks
> Yuan