SV: SV: SV: SV: SV: SV: SV: [Linux-cluster] GFS2 problem.

[Kristoffer Lippert]

Hi,

And thanks for all the posts. Not withstanding the fact that GFS2 to is
pre-production, i decided to spend a little more time on it.

I seem to have found another issue. 
The clusternodes (2 nodes alltogether), are connected to the San (FuSi
SX60 w. full redundancy) via. 2 port fibercontrollers. Thus i've set up
multipath, wich work fine. (tried pulling a cable, and it failed over
correctly).
But when i create much diskaccess from and to the san (ie. Copying a
100mb file from san to a new file on same partition) with the san
running GFS2, and the cluster suite running (RHEL5). I get a message in
the errorlog, saying that the lpfc just failed over. 
Why does it suddenly failover on the fiber, and does this smell a bit of
hardware issue to you too? 

If I do as much as a ls on the same volume from the other node, i get
the earlier mentioned error from GFS2, and the san is locked (I think
because of the fencing). Until i boot the faulted node.

Any sugestions or comments will be greatly appreciated,
Kristoffer
 

-----Oprindelig meddelelse-----
Fra: Brieseneck, Arne, VF-Group [mailto:Arne.Brieseneck at vodafone.com] 
Sendt: 4. juni 2007 17:06
Til: linux clustering; Kristoffer Lippert
Emne: RE: SV: SV: SV: SV: SV: SV: [Linux-cluster] GFS2 problem.

Hi David,

You know, that GFS1 is slow like hell... I truly can understand
Kristoffer when he wants to have GFS2. The throughput is much better.
On the other hand Kristoffer could use professional cluster filesystems
for PROD systems like Veritas FS. Afaik VFS is also available for RHEL.

Arne

-----Original Message-----
From: linux-cluster-bounces at redhat.com
[mailto:linux-cluster-bounces at redhat.com] On Behalf Of David Teigland
Sent: Montag, 4. Juni 2007 16:57
To: Kristoffer Lippert
Cc: linux-cluster at redhat.com
Subject: Re: SV: SV: SV: SV: SV: SV: [Linux-cluster] GFS2 problem.

On Mon, Jun 04, 2007 at 04:50:15PM +0200, Kristoffer Lippert wrote:
> Hi,
> 
> I've looked at the options, and it seems that building a new kernel is

> the only way to go about it.
> 
> RHEL 5.1 is scheduled for okt/nov, and that long we cannot wait.

GFS2 is still in preview/beta mode; you shouldn't use it for production
purposes yet (regardless of the kernel).  Just use GFS(1) instead.

Dave

--
Linux-cluster mailing list
Linux-cluster at redhat.com
https://www.redhat.com/mailman/listinfo/linux-cluster