Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 2006-05-29T17:25:07, Marcel Prisi <mprisi-lists at virtua.ch> wrote:
> I had a look at different things :
>
> - HaNFS / drbd-0.7 : needs 2 more servers, as I don't think having both
> the webservers and nfs/drbd on the same hardware makes any sense ??
> Would need to mount a local nfs export ?
> - drbd-8+GFS : seems really nice, but is it really ready for production ?
NFS back-end is likely the most stable approach, conservative and all
that.
drbd8 + OCFS2/GFS w/heartbeat 2.x is highly cutting edge, but I'm not
sure I'd want to run that ;-)
You could also run iSCSI backend servers (with drbd replicating across
them) and OCFS2 on top of that, which is slightly less dangerous than
the former, at least you get to use the stable DRBD version (and that'll
be supported with SLES10, at least for webserver workloads I think).
> - PeerFS : seems a good solution, but has anyone ever used it ? Is it
> stable ? Performance doesn't seem great at first glance ?
> - rsync : I need very low "convergence" ... no rsync ...
> - ddraid : seems really great ... ready for production ?
ddraid is unfinished code, I think.
> And what about these things my twisted brain imagined :
>
> - drbd / vblade / ata-over-ethernet / GFS or drbd / iSCSI / GFS ? Would
> it scale better than HaNFS ? Has anyone tried it ?
We've not done benchmarks. Probably it wouldn't - the bottleneck is the
network and the backend disks.
Sincerely,
Lars Marowsky-Brée
--
High Availability & Clustering
SUSE Labs, Research and Development
SUSE LINUX Products GmbH - A Novell Business -- Charles Darwin
"Ignorance more frequently begets confidence than does knowledge"