[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ssic-linux-users
Subject:    RE: [SSI-users] Re: openSSI webView 0.1 released!
From:       "Walker, Bruce J" <bruce.walker () hp ! com>
Date:       2004-11-02 15:10:38
Message-ID: 3689AF909D816446BA505D21F1461AE4C7525A () cacexc04 ! americas ! cpqcorp ! net
[Download RAW message or body]

Xavier,

To clarify, CFS does have caching.  It doesn't have "ondisk caching", which is what \
you propose.  I instead would suggest using DRBD to provide a consistent version of \
the root on two or more nodes (RH is also working on enhancements to CLVM to provide \
cross-node mirroring as well).  According the drbd web site the 2.4.x version is now \
stable so perhaps we should retry it, even before we go to the 2.6 kernel.

Bruce


> -----Original Message-----
> From: ssic-linux-users-admin@lists.sourceforge.net 
> [mailto:ssic-linux-users-admin@lists.sourceforge.net] On 
> Behalf Of Kilian CAVALOTTI
> Sent: Tuesday, November 02, 2004 4:03 AM
> To: En Chiang Lee
> Cc: Karl Vogel; 'kilian.cavalotti@stix.polytechnique.fr'; 
> Watson, Brian J.; ssic-linux-users@lists.sourceforge.net; 
> ssic-linux-devel@lists.sourceforge.net
> Subject: Re: [SSI-users] Re: openSSI webView 0.1 released!
> 
> 
> On Tuesday 02 November 2004 11:56, En Chiang Lee wrote:
> > > If you have shared disk hardware, the initnode can go down without
> > > the whole cluster going down.
> > 
> > The cluster will fail only when *all* the initnodes go down.
> 
> I know that. But with shared hardware, your init nodes will 
> very likely be 
> located in the same physical area, while your other nodes could be 
> wherever. So, in case of power (or network) outage, both of 
> your init nodes 
> will be unaccessible from the rest of the cluster, then, you 
> know what. :)
> 
> Well, all I wanted to say is that init nodes have a special 
> status on a 
> openSSI cluster, at least without a mean to dynamically 
> ditribute the root 
> filesystem all over the cluster. 
> 
> (Below are pure suggestions, with absolutly no idea of realizability)
> I think I've read CFS was based on NFS: can't we imagine a 
> version of CFS 
> with file caching, like AFS or CodaFS? This way, 'init nodes' 
> would become 
> 'booting nodes', and we could approach the real and perfect 
> peer-to-peer 
> cluster infrasctructure, with no special status for any node. 
> You probably 
> already know CacheFS, it can perhaps be useful in this case:
> http://www.kerneltraffic.org/kernel-traffic/kt20041030_281.html#2
> 
> -- 
> Kilian CAVALOTTI                               Ingénieur 
> Systèmes & Réseaux
> Laboratoire STIX                                        École 
> Polytechnique
> F91128 Palaiseau                                    Tel : +33 
> 1 69 33 41 13
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by:
> Sybase ASE Linux Express Edition - download now for FREE
> LinuxWorld Reader's Choice Award Winner for best database on Linux.
> http://ads.osdn.com/?ad_idU88&alloc_id065&op=ick
> _______________________________________________
> Ssic-linux-users mailing list
> Ssic-linux-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/ssic-linux-users
> 


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_idU88&alloc_id065&opÌk
_______________________________________________
Ssic-linux-users mailing list
Ssic-linux-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ssic-linux-users


[prev in list] [next in list] [prev in thread] [next in thread]