[Gluster-users] Taking glusterd peers down for maintenance

[Aaron Roberts]

Hi all,
	What is the best procedure to follow, to gracefully shutdown and then reintroduce a glusterd peer?

We are running a v3.1.0 4 peer Distributed-Replicate volume over TCP.  Each glusterd host provides 2 bricks (for a total of 8 bricks) and the bricks are ordered such that Replicate protects us against the loss of a host.  Each glusterd host also mounts the glusterfs volume using the FUSE client.  However, if we shutdown a host for maintenance, user processes running on the remaining hosts block on I/O to glusterfs for as much as 1 minute and then again for almost 5 minutes when the host rejoins the cluster.  How can we minimise this disruption?

Best regards,
  Aaron