[sheepdog-users] Cluster crash

[Hitoshi Mitake]

Hi Valerio,

At Fri, 6 Dec 2013 16:04:32 +0100,
Valerio Pachera wrote:
> 
> [1  <multipart/alternative (7bit)>]
> [1.1  <text/plain; UTF-8 (7bit)>]
> 2013/12/6 Valerio Pachera <sirio81 at gmail.com>
> 
> > I added the node named "sheepdog001" to my prouction cluster
> >
> ....
> >
> The cluster has crashed.
> >
> 
> I noticed I made a mistake: I run
> 
>   sheep -n  /var/sheep,/mnt/sheep/dsk01,/mnt/sheep/dsk02,/mnt/sheep/dsk03
> 
> instead of
> 
>   sheep -n  /var/sheep,/mnt/sheep/dsk01,/mnt/sheep/dsk02,/mnt/sheep/dsk03 \
>   -i host=192.168.5.41,port=3333
> 
> The other nodes are using a dedicated nic for sync.
> This node has stared with a single nic.
> In theory it should't have problems running with a single nic but somehting
> went wrong this time.
> 
> After the crash I killed all qemu, sheep, corosync processes on all nodes.
> I restarted corosync and sheep on the initial 3 nodes and then also on the
> new node (this time with both nics).
> 
> Everything is wokring fine.
> 
> It shouldn't matter but my sync nics are on a separated vlan.
> So it's like they are on a switch by their own.

Sorry, I can't solve this problem soon. Because we are not heavy users
of the IO nic feature.

Could you create a bug entry on our launchpad bug tracker?
https://bugs.launchpad.net/sheepdog-project

If you can add an issue with the detailed information to the tracker,
we would be able to reproduce your problem on our internal
environment.

Thanks,
Hitoshi