#619 Issue closed: Rear hangs upon restore for RHEL 7 for failure to mount /run

Labels: bug, support / question, fixed / solved / done

bibssun opened issue at 2015-07-13 18:30:

For RHEL 7 Server - rear restore appears to complete , but once the Server is rebooted - The Server hangs with the message "Failed to mount /run . No such file or directory."

bibssun commented at 2015-07-13 18:31:

The version of rear used is rear ver 1.17.

bibssun commented at 2015-07-13 18:35:

Also the method of recovery used is Rear Recovery with TSM .

gdha commented at 2015-07-14 08:33:

@bibssun could you check within TSM if /run was back-up'ed? Also, when you are still in recovery mode you could inspect the file systems under /mnt/local whether all required mount points are present.
You could verify the /var/lib/rear/layout/disklayout.conf file to see if /run was a separate mount point?

gdha commented at 2015-07-24 08:42:

On my centos7 system I have the following tmpfs file systems:

tmpfs                          12229932       92  12229840   1% /dev/shm
tmpfs                          12229932     9276  12220656   1% /run
tmpfs                          12229932        0  12229932   0% /sys/fs/cgroup

These are temporary file systems and are not tracked in the disklayout.conf file.

The problem with TSM is that variable MOUNTPOINTS_TO_RESTORE gets defined in script verify/TSM/default/40_verify_tsm.sh and afterwards as the variable is not empty in script restore/default/90_create_missing_directories.sh it will skip the creation of some critical mount points which are not backup'ed by TSM such as proc sys tmp dev/pts dev/shm.
Furthermore, I noticed that sys is missing in this list too.
Therefore, this is a bug, that needs to be fixed

gdha commented at 2015-07-24 08:50:

I think we need to do this in a more generic way and also restore the moint points with correct ownership and permissions. Therefore, we should make a new script in the prep work-flow and save the output for later and use it in the restore work-flow as input.

gdha commented at 2015-07-28 07:55:

A good starting point for the script could be:
$ ls -ld $(mount | grep -vE '(cgroup|fuse|nfs|/sys/)' | awk '{print $3}') to get the list of mount points where we need to save the permissions and ownership

gdha commented at 2015-08-13 08:13:

read some interesting comments from Schlomo in http://pikachu.3ti.be/pipermail/rear-users/2015-August/003138.html

gdha commented at 2015-11-04 20:37:

@bibssun feel free to test out the new script committed.

[Export of Github issue for rear/rear.]