st...@prd.co.uk (Steve Blinkhorn) writes: >Some off-list discussion has clarified matters. The fundamental problem is >that nfs >mounts are not restored automatically when an nfs server is rebooted - and that >may happen automatically so the sysadmin is unaware.
Actually that's the opposite from what NFS does. When the server comes back, the file operations will continue. I haven't seen anything else. >The connection with /etc/daily (etc.) is that find(1) hangs when it encounters >a broken >nfs mount point, gets stuck in tstile, and can't be killed. Here a find gets stalled in nfscn2 when the NFS server goes down. When mounted interruptibly (mount_nfs -i) you can kill the hanging operation, at least the hanging find. But interrupting is broken for other operations, e.g.: A read will keep the client process stuck non-interruptibly in uvnfp2, waiting for a page to become available. A write hangs in nfsaio which can be interrupted, but the process will then hang on exit as it waits for a synchronous close to succeed. The close operation requires the ioflush kernel thread to complete outstanding I/Os, and ioflush will be stuck in nfsrcv. The hanging ioflush will also impact all other I/O operations in the system.
