[DRBD-user] Slow NFS behaviour (0.7.5)

Bernd Schubert bernd-schubert at web.de
Tue Nov 23 00:50:18 CET 2004


On Monday 22 November 2004 23:47, you wrote:
> Hello Bernd,
>
> > which kernel are you using? Maybe you already found the thread about nfs
> > and 2.4. I started some time ago. With 2.6. drbd was running much faster
> > for us, but also even with failover the 2.6.7 was by far too unstable.
> > How do you mount the nfs export on your clients? If you will send your
> > problem to the NFS maillinglist, you will be first told to use tcp
> > mounts ;)
>
> You're going to eat me alive. 2.6.8.1. I know. I'm a cowboy :)

Well, we would also like using 2.6., it just didn't work for us. 

>
> Is the NFS over TCP the way to go then?

The linux NFS developers think that it should be used right now anyway. Its 
just not (yet) default, since other (non recent linux, etc) NFS servers might 
not support it yet.


>
> > Due to the instability of 2.6.7 we were forced to use 2.4.27 and we
> > certainly won't try 2.6.x in the near future on the server again. To
> > work around the slowness, we are also forced to risk some (hopefully
> > slight) dataloss:
> >
> > - unsing async NFS exports
> > - using DRBD protocol A
>
> I'm mounting sync. I'm afraid that since we need the NFS failover between
> the servers, A + async may not be an option for us :/ Our users would eat
> us alive.

Sure, we just didn't have much choice. With 2.4.27, sync-exports and protocol 
C it was just so slow, that normal working was not possible anymore :(
Fortunately the server is rock solid with 2.4.27 so we currently don't need 
failover.


Cheers,
 Bernd

-- 
Bernd Schubert
Physikalisch Chemisches Institut / Theoretische Chemie
Universität Heidelberg
INF 229
69120 Heidelberg
e-mail: bernd.schubert at pci.uni-heidelberg.de



More information about the drbd-user mailing list