Is this setup on bare metal or does it involve virtual machines?
I'm running GlusterFS on physical machines. No virtual machines involved.
Have you checked if port 24007 is reachable on all peers?
Yes, I did "nc -z server.xxx 24007" on all servers. All succeeded.
Additionally, if you are not running 3.7.3 you would need to ensure that the privileged ports check is bypassed in all glusterds.
# gluster volume get speech0 allow-insecure
gave me:
Option Value
------ -----
server.allow-insecure (null)
I'm not sure if the system's privileged ports are exausted.
Also I'm trying to set ping-timeout in /etc/glusterfs/glusterd.vol to a larger value, will this help?
Many thanks.
2015-09-01 16:18 GMT+08:00 Vijay Bellur <vbellur@xxxxxxxxxx>:
On Tuesday 01 September 2015 09:10 AM, Yiping Peng wrote:
Even if I'm seeing disconnected nodes (also from already-in-pool nodes),
my volume is still intact and available. So I'm guessing that glusterd
has few to do with volume/brick service?
Am I safe to kill all glusterd on all servers and start this whole peer
probing process all over again?
If I do this, will the currently mounted volumes become unavailable?
glusterd is the management daemon that takes care of cluster membership & configuration primarily.Normally a glusterd restart would not be needed to resolve the disconnected nodes' problem.
A few questions:
Is this setup on bare metal or does it involve virtual machines?
Have you checked if port 24007 is reachable on all peers?
Additionally, if you are not running 3.7.3 you would need to ensure that the privileged ports check is bypassed in all glusterds.
Regards,
Vijay
_______________________________________________ Gluster-users mailing list Gluster-users@xxxxxxxxxxx http://www.gluster.org/mailman/listinfo/gluster-users