Description of problem: Configured autofs to auto mount a remote NFS server mount point. The mount point is present in /etc/mtab and accessible from the command line. If the NFS server goes down and the mount point becomes stale/dead, any attempt to access the mount will result in a network timeout, then related error: # ls -al /mnt/autofs/users ls: /mnt/autofs/users: Permission denied IF you attempt to install an rpm while the mount is dead, rpm will throw an error: # rpm -Uvh wget-1.10.2-0.30E.i386.rpm --force error: failed to stat /mnt/autofs/users: Input/output error This is due to an over reaching RPM partition size test. RPM knows which files and directories it will attempt to write to during install, there's no need to test disk space for every partition on the server. This test is important, but inefficient and is the cause of many problems reported on linux forums all over the net, which remain un-answered. The --ignoresize parameter will skip the test and remove the error, but it defeats the point of the test in this case. Version-Release number of selected component (if applicable): rpm-4.2.3-24_nonptl How reproducible: Everytime. Steps to Reproduce: 1. Add any NFS mount point autofs or /etc/fstab. In /etc/auto.master: /mnt/autofs/ /etc/auto.ms --timeout=0 In /etc/auto.ms: users -rw,soft,retrans=4,intr,rsize=8192,wsize=8192 192.168.1.100:/var/spool/postfix/users Start up autofs and nfs respectively. Test that the NFS mount is accessible. 2. Stop the nfs daemon on the NFS server. 3. Test that access to the mount point times out. 4. Install any rpm. You can re-install the same rpm as well. Error should occur. Actual results: Mount point related error is displayed: # rpm -Uvh wget-1.10.2-0.30E.i386.rpm --force error: failed to stat /mnt/autofs/users: Input/output error Expected results: No error. I would expect RPM to not test unnecessary mounts for disk size. Additional info: The --ignoresize parameter should not be considered a work around. Also, the argument that system wide disk space checking is necessary because %pre and %post scripting might access partitions for which rpm files don't belong to isn't reasonable. NFS mount point shouldn't be tested unless the rpm package actually contains a file that is destined for said mount point.
Yes. This problem is fixed in rpm-4.4.8, as well as another obscure errno return from having a fuse-ssh mounted file system. UPSTREAM
FYI: It is necessary to gather size form all mount points, as symlinks may cross onto other mount points, and there is no way to know that a priori with a single pass install. Handling ESTALE is the usual issue however.
This bug is filed against RHEL 3, which is in maintenance phase. During the maintenance phase, only security errata and select mission critical bug fixes will be released for enterprise products. Since this bug does not meet that criteria, it is now being closed. For more information of the RHEL errata support policy, please visit: http://www.redhat.com/security/updates/errata/ If you feel this bug is indeed mission critical, please contact your support representative. You may be asked to provide detailed information on how this bug is affecting you.