Gluster users - Apr 2014 - NFS or Lustre like behavior when client/server becomes unreachable

Hi everyone,

I'm running Gluster 3.5.0 on small test cluster of 4 pcs. I've done some
testing
regarding fail-over and I'd like to ask about some of Glusters behavior.

When data nod or client is disconnected from network the running job 
(let's say writing with dd to some file)
on the client hangs for 42 (network.ping-timeout) and then recives error 
and ends. Is there any way to force
clients to wait infinate time until connection is reestablished? The 
same scenario is when data server goes down
due power loss etc.

What we would like to achieve is the same behavior as NFS or Lustre does 
where running client jobs hang until
the target is back online and then continues in the job.

Thanks for any ideas,
Michal Pazdera

---
Tato zpr?va neobsahuje viry ani jin? ?kodliv? k?d - avast! Antivirus je aktivn?.
http://www.avast.com

On Wed, 2014-04-30 at 21:29 +0200, Michal Pazdera wrote:
> What we would like to achieve is the same behavior as NFS or Lustre
> does 
> where running client jobs hang until
> the target is back online and then continues in the job.
This is what we would like too.

I think it's reasonable for a job that is in the process of writing a
file on a brick that disappears to fail but interrupted reads should be
recoverable.