I was "afraid" to delete anything since I have seen no DR docs related to this state. We did seem to reslove the situation however:
1. Did md5sum's of all file in /etc/glusterd and /etc/glusterfs on gbe02 and compared results to other 7 peers.
2. Noticed that the info file on gbe02 had a different Status then all others.
3. Noticed that the Status line in vols/peers/* differed on gbe02 while again... the other members had all identical files.
This entire problem occured while we were shutting down our servers for maintenance.
We knew there were no writes occuring during this time.
We decided to update gbe02's inof and vol/peers/* files to match the rest of the group and restart glusterd on it.
Wala... It joined the others.
We then did the self heal procedure on our gluster cluster and did a few tests to make sure that gbe02's writes were replicated to it's replica member.
We seem to be OK but I found it alarming there was no info on this out there.
What would you have done to handle this? Was there a simpler procedure?
From: Joe Landman <address@hidden>
Sent: Fri, May 6, 2011 6:54:34 PM
Subject: Re: [Gluster-devel] Gluster "Peer Rejected"
On 05/06/2011 09:48 PM, Nobodys Home wrote:
> Hello All,
> I have 8 servers.
> 7 of the 8 say that gbe02 is in state State: Peer Rejected (Connected).
> gbe08 says it is connected to the other 7 but they are all State: Peer
> Rejected (Connected)
> So it would appear that gbe02 is out of sync with the group.
Can the units ping each other? Could you delete the member that is out of sync, then re-initialize it and add it back in?
-- Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics, Inc.
web : http://scalableinformatics.com
phone: +1 734 786 8423 x121
fax : +1 866 888 3112
cell : +1 734 612 4615
Gluster-devel mailing listaddress@hiddenhttps://lists.nongnu.org/mailman/listinfo/gluster-devel