How to repair a file system corruption?

Question

Debian on external USB SSD drive. There was some error in dmesg log file:

...[    3.320718] EXT4-fs (sdb2): INFO: recovery required on readonly filesystem
[    3.320721] EXT4-fs (sdb2): write access will be enabled during recovery
[    5.366367] EXT4-fs (sdb2): orphan cleanup on readonly fs
[    5.366375] EXT4-fs (sdb2): ext4_orphan_cleanup: deleting unreferenced inode 6072
[    5.366426] EXT4-fs (sdb2): ext4_orphan_cleanup: deleting unreferenced inode 6071
[    5.366442] EXT4-fs (sdb2): 2 orphan inodes deleted
[    5.366444] EXT4-fs (sdb2): recovery complete
...

The system boots and works normally. Is it possible to repair this fully, and what is the proper way?

Solution 1

You can instruct the filesystem to perform an immediate fsck upon being mounted like so:

Method #1: Using /forcefsck

You can usually schedule a check at the next reboot like so:

$ sudo touch /forcefsck
$ sudo reboot

Method #2: Using shutdown

You can also tell the shutdown command to do so as well, via the -F switch:

$ sudo shutdown -rF now

NOTE: The first method is the most universal way to achieve this!

Method #3: Using tune2fs

You can also make use of tune2fs, which can set the parameters on the filesystem itself to force a check the next time a mount is attempted.

$ sudo tune2fs -l /dev/sda1
Mount count: 3
Maximum mount count: 25

So you have to place the "Mount count" higher than 25 with the following command:

$ sudo tune2fs -C 26 /dev/sda1

Check the value changed with tune2fs -l and then reboot!

NOTE: Of the 3 options I'd use tune2fs given it can deal with force checking any filesystem whether it's the primary's (/) or some other.

Additional notes

You'll typically see the "Maximum mount count:" and "check interval:" parameters associated with a partition that's been formatted as ext2/3/4. Often times they're configured like so:

$ tune2fs -l /dev/sda5 | grep -E "Mount count|Maximum mount|interval"
Mount count:              178
Maximum mount count:      -1
Check interval:           0 (<none>)

When the parameters are set this way, the device will never perform an fsck during mounting. This is fairly typical with most distros.

There are 2 forces that drive a check. Either number of mounts or an elapse time. The "Check interval" is the time based one. You can say every 2 weeks to that argument, 2w. See the tune2fs man page for more info.

NOTE: Also make sure to understand that tune2fs is a filesystem command, not a device command. So it doesn't work with just any old device, /dev/sda, unless there's an ext2/3/4 filesystem there, the command tune2fs is meaningless, it has to be used against a partition that's been formatted with one of those types of filessystems.

References

• Linux Force fsck on the Next Reboot or Boot Sequence

Solution 2

It is recommended to perform a fsck unmount this partition /dev/sdb2 if you have nothing important riding on that device, if that does not work try doing a fsck with a live CD.

$ sudo fsck /dev/sdb2

Comments

[Olivier Dulac]

If this happens after a brutal shutdown, no worries, it s probably tmp files. If it happened during use, then you should first backup everything important (on 2 separate supports, in case one fails)

[Lexx Luxx]

I already ran fsck -n -f check, it show no errors.

[user3132678]

problems and brings you this?

[Lexx Luxx]

the first two methods are one-time operations, and #3 allows set this operation permanently?

[slm]

Yeah it's set permanently, you generally need to go back and reset them after you're done with the fsck. At least that's what I do anyways.

[Lexx Luxx]

no problem while installation is still fresh, but I want to avoid an unexpected problems in the future. Was the file system completely fixed during recovery?

[user3132678]

That will depend on the status of the disk physically, checking typically linux hdd automates every any given time, so it can also be a chore, however you must be careful . I recommend you to run a fsck to dismount the partition at the time this is not writing or reading any device access as long as this is not the file system of your

[Lexx Luxx]

I have Mount count: 21 Maximum mount count: -1 Check interval:0 (<none>) It's not clear, what setting to specify?

[slm]

The "Max mnt cnt: -1" and "check interval: 0" are blocking any checks from ever occurring. make "Max mnt cnt" set to 20 and "check interval" can stay 0. There are 2 forces that drive a check. Either number of mounts or an elapse time. The 2nd argument is the time based one. You can say every 2 weeks to that argument, 2w. See the tune2fs man page for more info.

[Lexx Luxx]

these setting are used for partitions only, or can be for disk /dev/sdb?

[slm]

@triwo - see updates to A, hopefully they help.