6

I'm trying to shrink an ext4 filesystem on a CentOS 6 server. I did a lazy unmount of the filesystem while I waited for some processes to finish running. They've all finished running but I can't seem to do anything with the filesystem. How can I see what is using the volume and stop it?

resize2fs

[root@planck ~]# resize2fs -P /dev/vg_dev/lv_home 
resize2fs 1.42.9 (28-Dec-2013)
resize2fs: Device or resource busy while trying to open /dev/vg_dev/lv_home
Couldn't find valid filesystem superblock.

fsck

[root@planck ~]# fsck /dev/vg_dev/lv_home 
fsck from util-linux-ng 2.17.2
e2fsck 1.42.9 (28-Dec-2013)
/dev/mapper/vg_dev-lv_home is in use.
e2fsck: Cannot continue, aborting.

umount

[root@planck ~]# umount /dev/vg_dev/lv_home
umount: /dev/vg_dev/lv_home: not mounted

lvs

  LV      VG     Attr       LSize  Pool Origin Data%  Meta%  Move Log Cpy%Sync Convert
  lv_home vg_dev -wi-ao----  5.86t                                                    
  lv_root vg_dev -wi-ao---- 50.00g                                                    
  lv_swap vg_dev -wi-ao----  5.44g
Jayen
  • 1,827
  • 3
  • 16
  • 27
  • what is the output of `lvs`? – Aaron Feb 01 '16 at 10:24
  • @Aaron I added the output of lvs. looks the same as the mounted volumes. – Jayen Feb 02 '16 at 08:48
  • That suggests home is mounted. What do you get with `fuser -c /home` and `umount /home` and `grep home /etc/mtab` – Aaron Feb 02 '16 at 14:52
  • Sorry @Aaron, I've already rebooted. `fuser -mv /home` wasn't showing anything. `umount /home` gave the same `not mounted` error as `umount /dev/vg_dev/lv_home`, and `mount` did not show it as mounted (i assume `mtab` would have shown the same.) – Jayen Feb 03 '16 at 01:23

1 Answers1

9

fuser -mv /dev/vg_dev/lv_home should show you the process PID you need to kill to free up the device.

For example:

 # fuser -mv /dev/vg_dev/lv_home
                         USER        PID ACCESS COMMAND
    /dev/vg_dev/lv_home:
                         sbonds     9627 ..c.. bash

Now, you may not be able to actually kill it if it's blocked on I/O or some other uninterruptible reason.

Steve Bonds
  • 874
  • 2
  • 10
  • 19
  • 2
    no output :( nice to know `fuser` accepts a block device, though. – Jayen Feb 01 '16 at 08:33
  • 1
    One common non-process reason for a mount that won't move is another filesystem mounted on top, e.g. `/home/some-other-mount`. If that's not the case, you probably have a reboot in your near future. You need to [shrink ext4 offline anyhow](http://unix.stackexchange.com/questions/259182/why-the-ext4-filesystem-can-be-shrunk-only-when-not-mounted/259186#259186). – Steve Bonds Feb 02 '16 at 16:24
  • 1
    yeah there wasn't another system mounted on top. i rebooted, and kept killing processes until i could non-lazy unmount. `/home` is offline, but there are many services running out of `/var` that i can keep online during the resize. – Jayen Feb 03 '16 at 01:20
  • 1
    now I see the process:`/dev/dm-36: root kernel mount /media/snapshots/tmp-vm06.docker-disk` – but how do i [kill a process insde a snapshot?](https://serverfault.com/questions/926681/kill-a-process-inside-an-lvm-snapshot) – rubo77 Aug 16 '18 at 11:53
  • Something like "lsof" run globally may help you find the process(es) that have open files on that mount point. SystemTap may also assist. Pragmatically, usually the fastest way to resolve this is to reboot. :-) – Steve Bonds Aug 17 '18 at 14:14