Re: more information about CFS

Greg Freemyer <[email protected]> Mon, 1 Nov 2010 15:52:15 -0400
Newsgroups gmane.linux.cluster.ssic.user
Message-ID <[email protected]>
--===============8285699083746419352==
Content-Type: multipart/alternative; boundary=0022152d6e5955bdbc0494032603

--0022152d6e5955bdbc0494032603
Content-Type: text/plain; charset=ISO-8859-1

On Mon, Nov 1, 2010 at 12:53 PM, Alceu Rodrigues de Freitas Junior <
[email protected]> wrote:

> Hello guys,
>
> I'm looking for more information about CFS (Cluster File System) on
> OpenSSI.
>
> I already checked in the wiki, README.cfs and README.hardmounts but I'm
> looking for more introductory information like how does it works, if it
> uses NFS or not, etc.
>
> Could you please provide some more info about that? A link or two would
> be terrific.
>
> Thanks in advance,
> Alceu
>
>
>
I haven't worked on the CFS code myself, but I'm fairly sure it is NOT nfs
based.  ie. NFS is not used internally in the cluster to create a CFS
filesystem.

You can nfs export a CFS filesystem to computers outside the cluster, or a
openSSI cluster can mount externally provided NFS share.

Thus, NFS is a way for a openSSI cluster to work with the outside world, but
it is not used internally to the cluster itself.

CFS simply takes a local volume and makes it accessible as a filesystem in
parallel to all the nodes in the cluster.  For failover protection, you can
incorporate drbd.  drbd is not in the vanilla kernel code that openSSI is
compatible with.

(ie. It went into the vanilla kernel around 2.6.32 I think which is far
newer than the openSSI supported kernels).

Thus if you want CFS with failover on the cheap you should look at drbd.

You can also use CFS with externally provided raw volumes such as a FC SAN
or iSCSI SAN would provide.  In that case you need to have only one node at
a time accessing the raw volume.  It would in turn layer the CFS logic on
top and share the volume out to the other nodes as a filesystem.

You would need to implement node failover for the access/cfs/export
functionality to ensure the cluster could continue on in event of failure of
the node providing the connection to the raw volume.

Greg

--0022152d6e5955bdbc0494032603
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<br><br><div class=3D"gmail_quote">On Mon, Nov 1, 2010 at 12:53 PM, Alceu R=
odrigues de Freitas Junior <span dir=3D"ltr">&lt;<a href=3D"mailto:glasswal=
[email protected]">[email protected]</a>&gt;</span> wrote:<br><blockq=
uote class=3D"gmail_quote" style=3D"margin: 0pt 0pt 0pt 0.8ex; border-left:=
 1px solid rgb(204, 204, 204); padding-left: 1ex;">

Hello guys,<br>
<br>
I&#39;m looking for more information about CFS (Cluster File System) on<br>
OpenSSI.<br>
<br>
I already checked in the wiki, README.cfs and README.hardmounts but I&#39;m=
<br>
looking for more introductory information like how does it works, if it<br>
uses NFS or not, etc.<br>
<br>
Could you please provide some more info about that? A link or two would<br>
be terrific.<br>
<br>
Thanks in advance,<br>
Alceu<br>
<br>
<br></blockquote><div><br>I haven&#39;t worked on the CFS code myself, but =
I&#39;m fairly sure it is NOT nfs based.=A0 ie. NFS is not used internally =
in the cluster to create a CFS filesystem.<br><br>You can nfs export a CFS =
filesystem to computers outside the cluster, or a openSSI cluster can mount=
 externally provided NFS share.<br>

<br>Thus, NFS is a way for a openSSI cluster to work with the outside world=
, but it is not used internally to the cluster itself.<br><br>CFS simply ta=
kes a local volume and makes it accessible as a filesystem in parallel to a=
ll the nodes in the cluster.=A0 For failover protection, you can incorporat=
e drbd.=A0 drbd is not in the vanilla kernel code that openSSI is compatibl=
e with.<br>

<br>(ie. It went into the vanilla kernel around 2.6.32 I think which is far=
 newer than the openSSI supported kernels).<br><br>Thus if you want CFS wit=
h failover on the cheap you should look at drbd.<br><br>You can also use CF=
S with externally provided raw volumes such as a FC SAN or iSCSI SAN would =
provide.=A0 In that case you need to have only one node at a time accessing=
 the raw volume.=A0 It would in turn layer the CFS logic on top and share t=
he volume out to the other nodes as a filesystem.<br>

<br>You would need to implement node failover for the access/cfs/export fun=
ctionality to ensure the cluster could continue on in event of failure of t=
he node providing the connection to the raw volume.<br><br>Greg<br></div>

</div><br>

--0022152d6e5955bdbc0494032603--


--===============8285699083746419352==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

------------------------------------------------------------------------------
Nokia and AT&T present the 2010 Calling All Innovators-North America contest
Create new apps & games for the Nokia N8 for consumers in  U.S. and Canada
$10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing
Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store 
http://p.sf.net/sfu/nokia-dev2dev
--===============8285699083746419352==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Ssic-linux-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ssic-linux-users

--===============8285699083746419352==--