Daniel Gryniewicz

I got them, yes.

Sorry, I have upcoming vacation, and have things I need to get done for work. Maybe Frank can look in the interim, or I'll get to this when I'm back.

You don't need to tag me, I get all notifications for NTIRPC.

Is the traffic being delivered to the new ganesha instance? You can use the log (if logging is high enough), or a packet dump to determine this, I suspect. If...

I specifically meant traffic from the hung machine. If, for example, the switch didn't update it's ARP table for the port the client is on, or if the gratuitous ARP...

Several notes: - The connection to nfs1 will stay established until it times out. This is because the client can't terminate the connection. Usually, failover happens when the ganesha on...

The logging we're interested in is on the destination service, yes. What you propose should work. However, you don't need to restart Ganesha to change logging (or most config). You...

I'm not sure what's wrong, sessions are getting created correctly. Frank should be back tomorrow, maybe he has a better idea.

Locks must be preserved by the backend filesystem. Ganesha does not have clustered locking (and it couldn't be atomic if it did). The second ganesha instance will re-claim the lock...

Locks have to work cluster wide, but they do in FSAL_CEPH already.