NFS locking has always been fragile and assumed to be less than 100% reliable. I could be wrong, but it doesn't look like RHEL5 preserves NFS locks across a server reboot either. This makes it difficult to preserve them during a failover event which would be basically the same problem. Some systems have a way to make the lockd program (or equivalent) dump its state to disk during shutdown and load during startup, but I don't see that in the RHEL5 implementation.
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Collins, Kevin [Beeline] Sent: Friday, June 20, 2008 9:44 AM To: Red Hat Enterprise Linux 5 (Tikanga) discussion mailing-list Subject: RE: [rhelv5-list] Linux NAS cluster (RHCS/CTDB) So can I assume by the deafening silence that nobody out there is doing this stuff? I'm surprised... Thanks, Kevin ________________________________ From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Collins, Kevin [Beeline] Sent: Wednesday, June 18, 2008 10:34 AM To: [email protected] Subject: [rhelv5-list] Linux NAS cluster (RHCS/CTDB) Hi again, more questions related to clustering, specifically NFS and CIFS (Samba). My initial thoughts were that setting up the cluster would be the hard part and the NFS/CIFS piece would be relatively simple. Casual reads of RedHats literature says things like "built in support for NFS and Samba". I have read and re-read the docs, as well as scouring the net and I am finding that there is very little in the way of *actual* documentation on how to setup failover services for either using RedHat Cluster Suite. As of now, I have GFS filesystems which are mounted on both nodes of a 2-node cluster. The NFS daemon is started on both and both have identical /etc/exports. In the cluster, I have a Virtual IP resource and it is the sole resource of my NFS "service". This is a rather simple setup (once I figured it out) and I have confirmed that I can fail either node and have the virtual IP move to the other and NFS is still accessible. Next I was trying to figure out the CIFS piece of the puzzle, and started wondering how this would work with "security = domain" since the Samba server is joined to an AD domain. I couldn't figure out how the secret key would be managed between 2 (or more) nodes until I ran across someone mentioning moving that data to a GFS filesystem and then linking back to it. But then I thought, I can't have 2 active Samba nodes running sharing the same state data... In the course of searching, I also discovered (and found somewhere in the RedHat docs) that NFS locking is not maintained across a failover! So, I finally discovered CTDB (http://ctdb.samba.org <http://ctdb.samba.org> ) and it seems like it could be the solution to both the NFS and CIFS clustering concerns I have had. My questions are: 1) Is anyone really using RedHat Cluster Suite for failover NFS and/or CIFS? 1b) if so, can you provide some specifics on HOW you are doing so? 1c) does it adequately handle NFS file locking and state information? What about CIFS? 2) Is anyone using CTDB for failover NFS and/or CIFS? 2b) does it adequately handle NFS file locking and state information? What about CIFS? In addition, I would be happy to hear any additional opinions or input on your real-world experience with RHCS and/or CTDB, as well as any good reference sources (besides the manuals, which I already know of and have read). Thanks again - I relish having access to this forum where there is a wealth of opinion and knowledge. Kevin
_______________________________________________ rhelv5-list mailing list [email protected] https://www.redhat.com/mailman/listinfo/rhelv5-list
