LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re(2): DR question

To: Juri Haberland <juri.haberland@xxxxxxxxxxxxx>(Receipt Notification Requested), Juri Haberland <news-innominate.list.linux.lvs.users@xxxxxxxxxxxxx>(Receipt Notification Requested)
Subject: Re(2): DR question
Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx (Receipt Notification Requested)
From: SYLVAIN.COUTANT@xxxxxxxxxxx
Date: Tue, 1 Aug 2000 11:35:46 +0200
> SYLVAIN.COUTANT@xxxxxxxxxxx wrote:
> > 
> > > > Andy Bergeret wrote:
> > > >
> > > > i read the mon+heartbeat+fake+coda solution, but it's need more than
2
> > > > host ...
> > >
> > > It can be done with just two hosts.
> > 
> > Yes that's right, but the CODA solution does not seems to be the best. My
> own
> > tests:
> > 
> > - cp -r /etc/* /coda
> > 
> > That's about 16 Megs of files and it took more than 3 minutes to complete
> > with a venus cache of over 40 megs ... Awfully slow ... (tested with coda
> > 5.3.8)
> > 
> > CODA is really not ready for production purposes !
> > Beside this point, the technology is REALLY great ...
> 
> True. Instead of CODA i would suggest using drbd. It seems stable and is
> well maintained by Philipp Reisner. The other possibility is to use a
> shared SCSI system.

Yes, but DRBD only allows a two points mirroring (as far I know it). With A
3-4 computers cluster, this is not the best solution ...

Shared SCSI could be a good idea, but there are few systems wich would allow
more two connections. And the price is not the same ;-(( Will be the best
solution for systems with lots of write operations

The best thing in CODA is it allows more than one computer to mount the
filesystems. I do not know about any other technology allowing this (DRBD,
NBD, ...) This should be by far one of the greatest feature of CODA (with
support for disconnected operations).

Sylvain.

> 
> juri
> 
> -- 
> juri.haberland@xxxxxxxxxxxxx
> system engineer                                          innominate AG
> clustering & security                                networking people
> phone: +49-30-308806-45  fax: -77            web: http://innominate.de
> 
> 


<Prev in Thread] Current Thread [Next in Thread>