Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Yes, don't attempt to use the same servers to both serve NFS with DRBD
and mount the NFS volume. It is nothing but headaches. See
http://linux-ha.org/HaNFS
especially the Hints section.
- Dave
On Fri, Jun 12, 2009 at 09:26:15AM -0700, Stuart Sheldon wrote:
> Hi Keith,
>
> We actually did a similar setup using NFS. The only difference was we
> dedicated both DRBD servers to do nothing else but DRBD and NFS. We then
> use many (currently 8) front end servers that attach to the NFS share.
> This HOWTO was extremely helpful to me:
>
> http://www.howtoforge.com/high_availability_nfs_drbd_heartbeat
>
> Hope this helps.
>
> Stu
>
>
> Keith Edmunds wrote:
> > I'm new to the list. This has possibly been discussed before, in which
> > case a pointer would be welcome.
> >
> > We're planning a two-node heartbeat-controlled web server cluster possibly
> > using DRBD to share the web sites' data between the two nodes. Some of the
> > sites on this server are dynamically updated by users (eg, uploading
> > photos or whatever), so it's important that both web servers can write to
> > the shared data directories.
> >
> > Under consideration is using DRBD and making the primary accessible to the
> > other web server by mounting it r/w via NFS. I realise that the NFS
> > metadata will need to be on the DRBD volume, and that we need to get the
> > appropriate services stopped/started if there is a failover, but in
> > principle are there any problems with this? Is there any advice regarding
> > NFS parameters for such a configuration?
> >
> > We'd rather use ext3 + NFS than one of the cluster filesystems and a
> > master-master configuration, but if there are better ways of running a two
> > node webserver cluster where either node can update the data then we'd
> > like to consider them.
> >
> > Thanks for any help -
> > K.