Description of problem: Currently, even with device-mapper, there's a good chance that a device will have a different major/minor number on two different hosts. This causes problems with NFS exports because they use the device number to validate file handles on reboots. Rgmanager simultates reboots on failover, but unfortauntely, since the device might change between nodes, all of the filehandles will become invalid (stale). Fortunately, nfsd provides an export option, "fsid=x", were X is a symbolic file system ID which should be used instead of the device major/minor numbers in the file handle. On a failover, if the fsid matches, the client file handles are considered good irrespective of whether or not the actual device numbers match. Making this an inherit-able piece allows users to specify file system IDs as a property of the mount - rather than the individual client exports. This is what most people would want. Without this, nfsclient resources can not be reused; the user must create new ones for each service...
Created attachment 121998 [details] Implements fsid inheritance for fs/nfsexport/nfsclient/clusterfs
Patches not in CVS yet.
Patch in CVS - head, STABLE, RHEL4
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2006-0173.html