[LU-15167] fallocate does not increase quota usage

• Type: Bug
• Resolution: Fixed
• Priority: Major
• Fix Version/s: Lustre 2.15.0
• Severity: 3
• Assignee: Arshad Hussain
• Reporter: Andreas Dilger

Description

Using "fallocate -l <size> <file>" to allocate space to <file> does not increase the quota usage for the affected user/group/project:

# lfs quota -u 1000 /mnt/testfs
Disk quotas for usr 1000 (uid 1000):
     Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
    /mnt/testfs  102408       0       0       -       3       0       0       -
# lfs quota -g 1000 /mnt/testfs
Disk quotas for grp 1000 (gid 1000):
     Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
    /mnt/testfs  102408       0       0       -       3       0       0       -
# lfs quota -p 1000 /mnt/testfs
Disk quotas for prj 1000 (pid 1000):
     Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
    /mnt/testfs  102408       0  204800       -       3       0   10000       -

# fallocate -l 50M /mnt/testfs/projdir/50M
# lfs quota -u 1000 /mnt/testfs
Disk quotas for usr 1000 (uid 1000):
     Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
    /mnt/testfs  102408       0       0       -       4       0       0       -
# lfs quota -g 1000 /mnt/testfs
Disk quotas for grp 1000 (gid 1000):
     Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
    /mnt/testfs  102408       0       0       -       4       0       0       -
# lfs quota -p 1000 /mnt/testfs
Disk quotas for prj 1000 (pid 1000):
     Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
    /mnt/testfs  102408       0  204800       -       4       0   10000       -

Unmounting and remounting the whole filesystem does not help, I needed to run "e2fsck -fy" on the OSTs in order to update the quota file.

Issue Links

• is cloned by: LU-15519 fallocate does not increase projid quota usage
• is related to: LU-15748 interop: sanity test_150b: fallocate failed, error Operation not supported, mode 0, offset 62914560, len 4194304
• is related to: LU-14157 Implement fallocate() support for zfs Lustre
• is related to: LU-3606 Implement fallocate() support for ldiskfs Lustre

Comments

[Peter Jones]

Landed for 2.15

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/45475/
Subject: LU-15167 quota: fallocate send UID/GID for quota
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 789038c97ae1072874dcd6afcc9cd78e9c885368

[Gerrit Updater]

"Arshad Hussain <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/45475
Subject: LU-15167 quota: fallocate does not increase quota usage
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 50ec6d9d53748a3ecb8eb566a84dd200426f0cc3

[Arshad Hussain]

>... I needed to run "e2fsck -fy" on the OSTs in order to update the quota file.

I could not get this to work on my system. I thought I let you know. Maybe I am doing something wrong.

 

$ umount /mnt/lustre
$ umount /mnt/lustre-mds1
$ umount /mnt/lustre-ost1

$ umount /mnt/lustre-ost2
$ umount /mnt/lustre-ost3
$ e2fsck -fy /tmp/lustre-ost1

$ e2fsck -fy /tmp/lustre-ost2
$ e2fsck -fy /tmp/lustre-ost3

After remounting... Still no update.

$ lfs quota -u arshad /mnt/lustre
Disk quotas for usr arshad (uid 1000):
Filesystem kbytes quota limit grace files quota limit grace
/mnt/lustre 0 25600 25600 - 0 0 0 -

Sample e2fsck run...Done on all osts

 

$ e2fsck -fy /tmp/lustre-ost1
e2fsck 1.45.6.wc5 (09-Feb-2021)
Pass 1: Checking inodes, blocks, and sizes
Pass 2: Checking directory structure
Pass 3: Checking directory connectivity
Pass 4: Checking reference counts
Pass 5: Checking group summary information
[QUOTA WARNING] Usage inconsistent for ID 0:actual (2301952, 258) != expected (20480, 2)
Update quota info for quota type 2? yes


lustre-OST0000: ***** FILE SYSTEM WAS MODIFIED *****
lustre-OST0000: 268/100000 files (0.4% non-contiguous), 20245/100000 blocks

 

[Arshad Hussain]

Andreas, thanks for pointing out. I could recreate it in my setup. I am looking into this.

>This is probably ten times as important to fix for ZFS as it is for ldiskfs, because there isn't really a "quota repair" tool for zfs like e2fsck for ldiskfs.

Understood. Will handle this in the ZFS implementation also.

Fail Case:

$ fallocate -l 1M /mnt/lustre/t1
$ lfs quota -u arshad /mnt/lustre
Disk quotas for usr arshad (uid 1000):
Filesystem kbytes quota limit grace files quota limit grace
/mnt/lustre 0 25600 25600 - 0 0 0 -

Pass Case:

$ dd if=/dev/zero of=/mnt/lustre/t7 bs=2M count=1
$ lfs quota -u arshad /mnt/lustre
Disk quotas for usr arshad (uid 1000):
Filesystem kbytes quota limit grace files quota limit grace
/mnt/lustre 2048 25600 25600 - 2 0 0 -
uid 1000 is using default file quota setting

 

[Andreas Dilger]

This is probably ten times as important to fix for ZFS as it is for ldiskfs, because there isn't really a "quota repair" tool for zfs like e2fsck for ldiskfs.