Hello, I configured a Linux-HA-Cluster of two nodes for a redundant NFS Service. Its active/passive configuration and do not use DRBD.
In my configuration I manage data coming from a raid array, mirrored with some MD Devices (software mirrors), distributed over a lot of file systems (xfs+lvm) and then exported by NFS. Linux-HA starting point is a self written ressource for the MD devices, followed by OCF LVM ressources. Then every filesystem mount is a ressource in linux-ha (from OCF Filesystem RA). I decided to set an order condition from every file system mount to the ressource of the NFS Server. So, NFS Server is not starting before any filesystem is mounted. But seems this is not a good idea, because if any of the file systems fails, the whole NFS Service breaks down (not started or would be stopped by heartbeat because of the ordering). Can anyone give me a suggestion how to set up the ordering of the ressources? Should NFS server starts without any dependency or should it start after md devices or LVM rather then after mounting the file systems? I also think about writing a new OCF compliant RA, which mounts all my file systems at once, not only one per ressource. An attribute of such a RA could be the name of a config file similar to /etc/fstab with all necessary file systems and its devices, mount points, mount options etc. But maybe it is better to use OCF Filesystem RA furthermore and set up a ressource group then hidden the process of mounting file systems within one script. Any opinion? Thanks in advance for help, Annette _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
