[Gluster-devel] How to get rid of NFS on 3.7.0?

[Atin Mukherjee]

On 05/18/2015 08:50 PM, Atin Mukherjee wrote:
> Here is the issue:
> 
> Locking on a volume fails with the following error:
> 
> [2015-05-18 09:47:56.038463] E
> [glusterd-syncop.c:562:_gd_syncop_mgmt_lock_cbk] 0-management: Could not
> find peer with ID 70e65fb9-cc9d-16ba-a4f4-5fb901000000
> 
> [2015-05-18 09:47:56.038527] E [glusterd-syncop.c:111:gd_collate_errors]
> 0-: Locking failed   on 85eb78cd-8ffa-49ca-b3e7-d5030bc3124d. Please
> check log file for details.
> [2015-05-18 09:47:56.038574] E
> [glusterd-syncop.c:1804:gd_sync_task_begin] 0-management:      Locking
> Peers Failed.
> 
> From the above log it is clear that peer was not found and that's
> because of an incorrect peer id passed to glusterd_peerinfo_find
> 
> http://review.gluster.org/#/c/10192 has introduced this problem. In
> _gd_syncop_mgmt_lock_cbk peerid is taken from the frame->cookie which is
> an address of a local variable in gd_syncop_mgmt_lock () which means the
> moment this function goes out of scope peerid is not the same. We would
> need to have pointer to peerid and allocate/deallocate based on demand
> to solve it (more or less in similar to
> http://review.gluster.org/#/c/10192/1 )
Posted a patch [1] to fix this. I'll backport it to 3.7 once its
accepted in mainline.

[1] http://review.gluster.org/#/c/10842/

~Atin
> 
> ~Atin
> 
> On 05/18/2015 02:15 PM, Emmanuel Dreyfus wrote:
>> On Mon, May 18, 2015 at 01:48:52AM -0400, Krishnan Parthasarathi wrote:
>>> I am not sure why volume-status isn't working.
>>
>> My understanding is that glusterd considers a lock is held by the 
>> NFS comonent, while it is not started.
>>
> 

-- 
~Atin