[Gluster-devel] deadlock? Unable to get lock for uuid which is held by itself

[Yandong Yao]

Thanks Vijay for your response. My env has 2 nodes gluster cluster, and I
use replica=2 for a volume.  On another node (actually a laptop), I have
two scripts, one script will continuously connect to one node of the
gluster cluster, and do some read/write operations, if it could not connect
to this node, the script will try another node automatically.  Another
script will login to one node and performing random reboot operations to
simulate node down.

After about 1 hour, I got above issue.

Regards,
Yandong


2014-01-27 Vijay Bellur <vbellur at redhat.com>

> On 01/25/2014 06:27 PM, Yandong Yao wrote:
>
>> Hi Guys,
>>
>> I am using gluster 3.4.2 and doing some test with replica=4, and got
>> following 'unable to get lock error', while it seems the lock has been
>> acquired by itself already. is this a deadlock?
>>
>> any other volume command will be fail with error message like 'another
>> transaction is in progress'.
>>
>
> The locking problem is related to the cluster wide locking done by
> glusterd for performing a volume operation. A restart of glusterd on all
> nodes can help in overcoming this problem.
>
> Is there a sequence of steps that lead to this problem?
>
>
> -Vijay
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20140128/a1c420b6/attachment-0001.html>