I have this scenario:
-
2 NFS servers with /usr/share/man as exports (just as example, it can be anything), let say server A and server B (not replicas, just 2 independent servers on the same subnet sharing the same things)
-
both NFS servers are up and healthy
-
user is initiating mount from client by executing ls /mnt/usr/man (as example); timeout in auto.master is 2 minutes, fstype=nfs, options=ro,soft
-
automount is successful, server A was chosen by autofs, and user can see listed man pages
-
after 1m 10sec server (NFS) A goes down
-
at 1m 20 sec user executes ls /mnt/usr/man again
What is happening is that it is locked – mount stays on server A and no user can go into /mnt/usr/man (nor lsof is working)
After waiting some minutes (ca 10 minutes)
ls: cannot open directory /mnt/usr/man: Input/output error
My question is: should autofs manage this scenario and automatically remount to server B? (even if mount timeout has not finished yet?)
I have read many articles and it looks like my scenario is not "autofs"'es duty (If I want something like this it should be done another way).
But I want to be sure 100% whether autofs should handle this situation itself or not.
Best Answer
I know this is pretty late to the party, but was the nfs-server service started on server B? I just ran into these exact symptoms, and that was the solution. Try running
service nfs-server start
on server B.