[LU-15229] conf-sanity test_78: FAIL: shrink mds1 to 307837 failed

• Status: Open
• Project: Lustre
• Affects Version/s: Lustre 2.15.0
• Type: Bug
• Priority: Minor
• Reporter: Elena Gryaznova
• Assignee: WC Triage
• Resolution: Unresolved
• Severity: 3

Description

         219 regular files
         139 directories
           0 character device files
           0 block device files
           0 fifos
           0 links
           0 symbolic links (0 fast symbolic links)
           0 sockets
------------
         358 files
Memory used: 272k/0k (82k/191k), time: 11.09/ 0.06/ 0.01
I/O read: 6MB, write: 1MB, rate: 0.54MB/s
lm0321: resize2fs 1.45.6.cr1 (14-Aug-2020)
lm0321: resize2fs: MMP: e2fsck being run while trying to open /dev/vdb
pdsh@lm0327: lm0321: ssh exited with exit code 1
Couldn't find valid filesystem superblock.

Comments

[Andreas Dilger]

Per SEA-77:

It turns out resize2fs doesn't reset the MMP block back to the clean state after running it on a device that has MMP feature enabled:

# dumpe2fs -h /dev/disk/by-id/scsi-1IET_00020001 | grep mmp
dumpe2fs 1.42.13.wc5 (15-Apr-2016)
Filesystem features:      has_journal ext_attr resize_inode dir_index filetype mmp flex_bg dirdata sparse_super large_file huge_file uninit_bg dir_nlink quota

# resize2fs -P /dev/disk/by-id/scsi-1IET_00020001
resize2fs 1.42.13.wc5 (15-Apr-2016)
Estimated minimum size of the filesystem: 80788

# resize2fs -P /dev/disk/by-id/scsi-1IET_00020001
resize2fs 1.42.13.wc5 (15-Apr-2016)
resize2fs: MMP: fsck being run while trying to open /dev/disk/by-id/scsi-1IET_00020001
Couldn't find valid filesystem superblock.

So, this is a defect in resize2fs. I'll investigate how to resolve the issue.

A workaround is to run the following command after running resize2fs:

# tune2fs -f -E clear-mmp /dev/disk/by-id/scsi-1IET_00020001