Hey,
Has anyone had any success with deploying an Active/Active NFS server? I'm curious how lock recovery is performed.
In a typically Active/Passive scenario we have a nfs-server instance coupled with the exportfs. The nfs lock info is stored on some shared storage that follows that nfs server and the exportfs instances around the cluster. This allows us to alert the nfs clients after the failover that the server rebooted and that they need to re-establish their locks.
With an Active/Active setup, we'd have multiple nfs servers and exportfs instances, non of which are tied to one another. Meaning that the exportfs resources could run on any of the nfs server instances within the cluster. On failover, if we wanted the exportfs resources on a failed node to be taken over by another already existing nfs server on another node. In this instance does anyone know of a good way to alert the nfs clients previously connected to the old (failed) node that they need to re-establish their locks with the new node? It seems like the statd info from both the failed node's nfs server and the new node's nfs server would have to be merged or something.
any thoughts?
-- Vossel
_______________________________________________
Linux-HA mailing list
Linux-HA@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
Has anyone had any success with deploying an Active/Active NFS server? I'm curious how lock recovery is performed.
In a typically Active/Passive scenario we have a nfs-server instance coupled with the exportfs. The nfs lock info is stored on some shared storage that follows that nfs server and the exportfs instances around the cluster. This allows us to alert the nfs clients after the failover that the server rebooted and that they need to re-establish their locks.
With an Active/Active setup, we'd have multiple nfs servers and exportfs instances, non of which are tied to one another. Meaning that the exportfs resources could run on any of the nfs server instances within the cluster. On failover, if we wanted the exportfs resources on a failed node to be taken over by another already existing nfs server on another node. In this instance does anyone know of a good way to alert the nfs clients previously connected to the old (failed) node that they need to re-establish their locks with the new node? It seems like the statd info from both the failed node's nfs server and the new node's nfs server would have to be merged or something.
any thoughts?
-- Vossel
_______________________________________________
Linux-HA mailing list
Linux-HA@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems