[Linux-cluster] Please correct me if I'm wrong, but...
dmp at mjm.com
Mon Aug 20 20:15:03 UTC 2007
My application is slightly different, but I went through all of this very recently, and depending on a couple of variables, the answer may be; "no, you do not need GFS."
I am in the process of configuring a two-node cluster, with each acting as an NFS server. In the event of a node failure, the surviving node takes ownership of the file systems and exports them as NFS shares. Initially, we built the systems with RHEL 4 Update 4 and Cluster Suite. Our initial attempts to build the NFS services did no go well because of problems with this version of Cluster Suite's NFS facility. Red Hat support suggested going back to RHEL 4 Update 3, which was not an option for us, so we built a partially manual failover.
But then RHEL4 Update 5 and RHEL5 were released, and we now had two options:
1 - Update to the RHEL4 U5 release to configure service with NFS as it was originally designed. According to Red Hat support, the Cluster Suite NFS "bug" introduced with the RHEL4 Update 4 release was fixed in RHEL4 Update 5, but not RHEL5.
2 - Update to RHEL5, which repackages the whole Cluster Suite product with GFS bundled into it at no additional cost. Prior to this, GFS had to be purchased. This would have added a layer of complexity to the configuration we don't need at this time.
You may want to wait for someone else to address the "continuity" component of your question.
I hope this helps.
P.S.--According to the documentation, the RHEL4 Update 3 release had a working "NFS" facility and a utility to import the NFS configuration from the existing /etc/exports file. Unfortunately, this was dropped with RHEL4 Update 4 and not restored with Update 5. This would have made this build of this configuration so much easier. Oh well.
From: linux-cluster-bounces at redhat.com on behalf of Randy Brown
Sent: Mon 8/20/2007 15:26
To: linux clustering
Subject: [Linux-cluster] Please correct me if I'm wrong, but...
in order to configure a two-node high availability NFS failover cluster,
I need to use GFS, correct?
I am wanting to configure two machines in a cluster and use them as a
NAS head for an ISCSI based storage unit providing NFS file systems to
the machines on our network. I'd like to have the units configured so
if one fails, the other takes over and no interruption to the nfs mounts
(high-availability failover). Everything I'm reading indicates that GFS
would be required to do this, but would like confirmation. In case I'm
just missing something. Worst case, I guess I could just have our
second machine set up and ready to go as a warm spare.
Any thoughts or suggestions would be greatly appreciated.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 6055 bytes
Desc: not available
More information about the Linux-cluster