[sheepdog-users] strange behavior when deconfiguring a nic

[Liu Yuan]

On 12/21/2012 05:32 PM, Valerio Pachera wrote:
> I posted this isssue before but I think it deserves a thread it self.
> 
> *If I deconfigure the nic of a node, it's impossible to get back the
> node in the cluster.*
> 
> On the 3th node of my cluster:
>   ifdwon eth0
> or
>   ip addr del 192.168.2.43/24 dev eth0
> 
> Once restored the ip, I run sheep daemon again.
> It starts and in sheep.log you can see this:
>   Dec 21 10:17:18 [main] jrnl_recover(230) opening the directory
> /mnt/sheepdog/journal/
>   Dec 21 10:17:18 [main] jrnl_recover(235) starting journal recovery
>   Dec 21 10:17:18 [main] jrnl_recover(291) journal recovery complete
>   Dec 21 10:17:18 [main] send_join_request(1014) IPv4 ip:192.168.2.43 port:7000
>   Dec 21 10:17:19 [main] main(616) sheepdog daemon (version
> 0.5.5_23_gfacdf48) started
>   Dec 21 10:17:21 [main] update_cluster_info(799) status = 4, epoch =
> 8, finished: 0
> 
> Node 1 and node 2 don't show node3 by 'cluster info'.
> Node 3 shows all nodes by 'cluster info' (the state of the cluster
> before it was going off)
> The node recovery do not start.
> 
> To kill and restart sheep daemon doesn't help.
> I also tried to restart corosync without luck.
> *After a reboot of the node, the sheep daemon is able to join the cluster!*
> 
> More details in the attached file.

Maybe before restarting the sheep daemon, you'd better restart the
corosync daemon first. Please try if this solve the problem.

Thanks,
Yuan