[Gluster-users] Glusterfs client stops

Stephan von Krawczynski skraw at ithnet.com
Tue Jun 16 11:53:23 UTC 2009


On Tue, 16 Jun 2009 06:35:56 -0500 (CDT)
Vikas Gorur <vikas at gluster.com> wrote:

> 
> ----- "Stephan von Krawczynski" <skraw at ithnet.com> wrote:
>  
> > Hopefully you do not disconnect after one lost "ping". Unfortunately
> > one can
> > experience some lost packets over the day. There is few you can do
> > about it.
> > Some are lost because of real disconnects but most are just glitches
> > in
> > switches and network drivers. Are these real tcp
> > disconnects/timeouts?
> 
> We disconnect after one lost ping because otherwise a command like "ls" or
> an application could hang forever. Reconnect is attempted soon after the disconnect.
> 
> Vikas

But disconnect means that a working application (like some php app on a
webserver) will most likely crash on the user because of missing files,
right?. If it only took some time to respond he may be a lot happier.
"ls" btw can hang around quite some time anyway because you are syncing the
files on servers nodes on such an event (in case they are out of sync) and
moving a lot of data in the background...
nfs hangs, too, (trying to reconnect) as far as I remember ...
-- 
Regards,
Stephan




More information about the Gluster-users mailing list