NFS and Non-Responsiveness

[ginoitalo]

I have 4 webservers serving content from an NFS mount on a RAID5 box. All FreeBSD.

NFS Server:
----------------------------
nfs_server_enable="YES"
nfs_server_flags="-u -t -n 8"
rpc_lockd_enable="YES" # Run NFS rpc.lockd needed for client/server.
rpc_statd_enable="YES" # Run NFS rpc.statd needed for client/server.
nfs_client_enable="YES" # This host is an NFS client (or NO).
nfs_reserved_port_only="NO"
mountd_flags="-r"


Webserver mounts using the fstab file


Issue I recently had:
------------------------

The NFS server went down.
I expect the webservers to not be able to serve content, but they all became non-usable as soon as i typed one of the following:

df -h
umount
umount -f
reboot

...couldn't even reboot for crying out loud, network services turned off, but it wouldn't go all the way, so i had to drive to the datacenter and hard reboot.


anyways, I can't imagine that a 'bad share' could take down such a great OS.


Is there anything I can do to prevent this hangup in the future ?

[Fordy]

How about holding a local cache on each server. How you did this could be up to you. It could be something simple like a web directory that gets rsync'ed at a regular interval. Then if the main server goes down the sync fails but the webservers keep going with the cached content.

I doubt a failed share serving only web content should kill the system though