From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20020724 Description of problem: Directories are mounted on a RedHat Linux client system from an SGI Irix NFS server. The mounts are automatically handled by amd. If the server is rebooted, the mount points sometimes go into some kind of error state where attempting to read from anything mounted there gives an Input/output error, attempting to umount gives a busy error, and attempting to use fuser to determine what processes are using the mounted directory gives an Input/output error. It seems that the only way to clear the error is to reboot the client machine. Version-Release number of selected component (if applicable): How reproducible: Sometimes Steps to Reproduce: I believe that in order to reproduce, one or more files on the mounted NFS filesystem must be open when the NFS server is rebooted. It is not possible to reproduce this problem reliably for any one mount point, but if our NFS server is rebooted, it is almost inevitable that at least one of our 4 heavily used client machines will need to be rebooted. Actual Results: Attempting to access any file in the NFS mounted directory gives Input/output error. Expected Results: Access to the NFS mounted directory should work normally once the NFS server is back up and running. Additional info: The machines in question have been updated to all the latest errata, and while the problem most frequently occurs on the one running 7.2, it also occurs on machines running 7.3. It is probable that the machine most commonly affected is not because of the different version, but because it makes heavier use of NFS mounted filesystems.
I'm having a similar issue here. My problem is the client is in fact the server which mounts some directories in order to share config files from a mutual nfs mount. What I would find useful would be that nfs looked up the hostname each times it accesses the nfs server, since in this case srv1.domain.dyn might switch IP addresses sometimes. Mar 28 01:30:42 cents kernel: nfs: server srv1.domain.dyn not responding, timed out Mar 28 01:40:21 cents last message repeated 2 times Mar 28 01:40:21 cents kernel: nfs_statfs: statfs error = 5 Mar 28 01:40:42 cents kernel: nfs: server srv1.domain.dyn not responding, timed out Mar 28 01:41:41 cents kernel: nfs: server srv1.domain.dyn not responding, timed out I've got the same remarks about umount (-f) and fuser as the reporter of this bug.
I'm seeing this on Everest (ia64) running RHEL3 beta2. # cd /store/ISO/redhat/taroon/ia64/ # ll ls: taroon-SRPMS-beta2-ia64-disc3.iso: Input/output error ls: MD5SUMS: Input/output error ls: taroon-rhel3as-beta2-ia64-disc2.iso: Input/output error ls: taroon-rhel3as-beta2-ia64-disc3.iso: Input/output error ls: taroon-rhel3as-beta2-ia64-disc4.iso: Input/output error ls: taroon-SRPMS-beta2-ia64-disc1.iso: Input/output error ls: taroon-rhel3as-beta2-ia64-disc1.iso: Input/output error ls: taroon-SRPMS-beta2-ia64-disc2.iso: Input/output error The NFS server describes itself this way: $ ssh test.lsy uname -a Linux linuxtest.fc.hp.com 2.4.19-686-smp #1 SMP Tue Nov 19 00:58:50 EST 2002 i686 Pentium III (Cascades) GenuineIntel GNU/Linux $
This seems to work on later kernels.