[LU-5028] quota has grace time when the soft limits are not exceeded Created: 08/May/14 Updated: 18/Nov/16 Resolved: 18/Nov/16 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.4.2 |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Critical |
| Reporter: | Li Xi (Inactive) | Assignee: | Niu Yawei (Inactive) |
| Resolution: | Cannot Reproduce | Votes: | 0 |
| Labels: | None | ||
| Attachments: |
|
| Severity: | 3 |
| Rank (Obsolete): | 13909 |
| Description |
|
We found 'lfs quota' reports strange results for some groups. The soft limit of them are not exceeded, but the grace times are not zero. [root@fm08 ~]# lfs quota -g gism /home2 [root@fm08 ~]# lfs quota -g gitj /home2 |
| Comments |
| Comment by Niu Yawei (Inactive) [ 08/May/14 ] |
|
I'm afraid that the allocated limits has exceeded the softlimit, could you run "lfs quota -v" to see if the total allocated block limit has exceeded the softlimit? |
| Comment by Li Xi (Inactive) [ 08/May/14 ] |
|
Hi Yawei, Thanks for the quick response! Following is the 'lfs quota -v' results. We changed the softlimit from 200000000 to 1000000000, so there is no grace time now. [root@ff04 ~]# lfs quota -v -g gism /home2
Disk quotas for group gism (gid 1513):
Filesystem kbytes quota limit grace files quota limit grace
/home2 45897736 1000000000 2000000000 - 1 0 0 -
home2-MDT0000_UUID
4 - 0 - 1 - 0 -
home2-OST0000_UUID
256004 - 16777216 - - - - -
home2-OST0001_UUID
0 - 4194384 - - - - -
home2-OST0002_UUID
0 - 4718844 - - - - -
home2-OST0003_UUID
256000 - 16777216 - - - - -
home2-OST0004_UUID
0 - 5242980 - - - - -
home2-OST0005_UUID
2676740 - 16777216 - - - - -
home2-OST0006_UUID
2676740 - 16777216 - - - - -
home2-OST0007_UUID
2676740 - 16777216 - - - - -
home2-OST0008_UUID
2932744 - 16777216 - - - - -
home2-OST0009_UUID
0 - 4788916 - - - - -
home2-OST000a_UUID
256004 - 16777216 - - - - -
home2-OST000b_UUID
0 - 4194304 - - - - -
home2-OST000c_UUID
0 - 4194304 - - - - -
home2-OST000d_UUID
256000 - 16777216 - - - - -
home2-OST000e_UUID
0 - 4194304 - - - - -
home2-OST000f_UUID
2676740 - 16777216 - - - - -
home2-OST0010_UUID
2676740 - 16777216 - - - - -
home2-OST0011_UUID
2676740 - 16777216 - - - - -
home2-OST0012_UUID
2932744 - 16777216 - - - - -
home2-OST0013_UUID
0 - 4823972 - - - - -
home2-OST0014_UUID
0 - 4194428 - - - - -
home2-OST0015_UUID
0 - 4195184 - - - - -
home2-OST0016_UUID
0 - 4719436 - - - - -
home2-OST0017_UUID
256004 - 16777216 - - - - -
home2-OST0018_UUID
0 - 0 - - - - -
home2-OST0019_UUID
2676204 - 16777216 - - - - -
home2-OST001a_UUID
2676204 - 16777216 - - - - -
home2-OST001b_UUID
2676204 - 16777216 - - - - -
home2-OST001c_UUID
2932208 - 16777216 - - - - -
home2-OST001d_UUID
256004 - 16777216 - - - - -
home2-OST001e_UUID
0 - 4194808 - - - - -
home2-OST001f_UUID
0 - 4718896 - - - - -
home2-OST0020_UUID
256000 - 16777216 - - - - -
home2-OST0021_UUID
0 - 5242892 - - - - -
home2-OST0022_UUID
2676740 - 16777216 - - - - -
home2-OST0023_UUID
2676740 - 16777216 - - - - -
home2-OST0024_UUID
2676740 - 16777216 - - - - -
home2-OST0025_UUID
2932744 - 16777216 - - - - -
home2-OST0026_UUID
0 - 4543924 - - - - -
home2-OST0027_UUID
256004 - 16777216 - - - - -
[root@ff04 ~]# lfs quota -v -g gitj /home2
Disk quotas for group gitj (gid 3271):
Filesystem kbytes quota limit grace files quota limit grace
/home2 16644836 1000000000 2000000000 - 1 0 0 -
home2-MDT0000_UUID
4 - 0 - 1 - 0 -
home2-OST0000_UUID
252836 - 16777216 - - - - -
home2-OST0001_UUID
232308 - 16777216 - - - - -
home2-OST0002_UUID
304368 - 16777216 - - - - -
home2-OST0003_UUID
479444 - 16777216 - - - - -
home2-OST0004_UUID
304508 - 16777216 - - - - -
home2-OST0005_UUID
237444 - 16777216 - - - - -
home2-OST0006_UUID
1175356 - 16777216 - - - - -
home2-OST0007_UUID
239768 - 16777216 - - - - -
home2-OST0008_UUID
685248 - 16777216 - - - - -
home2-OST0009_UUID
243156 - 16777216 - - - - -
home2-OST000a_UUID
275252 - 16777216 - - - - -
home2-OST000b_UUID
227748 - 16777216 - - - - -
home2-OST000c_UUID
311196 - 16777216 - - - - -
home2-OST000d_UUID
511500 - 16777216 - - - - -
home2-OST000e_UUID
293564 - 16777216 - - - - -
home2-OST000f_UUID
218748 - 16777216 - - - - -
home2-OST0010_UUID
1186936 - 16777216 - - - - -
home2-OST0011_UUID
444848 - 16777216 - - - - -
home2-OST0012_UUID
504152 - 16777216 - - - - -
home2-OST0013_UUID
229696 - 16777216 - - - - -
home2-OST0014_UUID
245656 - 16777216 - - - - -
home2-OST0015_UUID
272548 - 16777216 - - - - -
home2-OST0016_UUID
457932 - 16777216 - - - - -
home2-OST0017_UUID
359204 - 16777216 - - - - -
home2-OST0018_UUID
234288 - 16777216 - - - - -
home2-OST0019_UUID
223420 - 16777216 - - - - -
home2-OST001a_UUID
1166492 - 16777216 - - - - -
home2-OST001b_UUID
440788 - 16777216 - - - - -
home2-OST001c_UUID
491780 - 16777216 - - - - -
home2-OST001d_UUID
250044 - 16777216 - - - - -
home2-OST001e_UUID
245992 - 16777216 - - - - -
home2-OST001f_UUID
273804 - 16777216 - - - - -
home2-OST0020_UUID
466008 - 16777216 - - - - -
home2-OST0021_UUID
348564 - 16777216 - - - - -
home2-OST0022_UUID
239476 - 16777216 - - - - -
home2-OST0023_UUID
229744 - 16777216 - - - - -
home2-OST0024_UUID
1172008 - 16777216 - - - - -
home2-OST0025_UUID
448984 - 16777216 - - - - -
home2-OST0026_UUID
473704 - 16777216 - - - - -
home2-OST0027_UUID
246320 - 16777216 - - - - -
|
| Comment by Niu Yawei (Inactive) [ 08/May/14 ] |
|
If you change the softlimit, the allocated limits could be changed as well. (because the qunit is increased when you enlarge the softlimit) Did you unlink some large files before? (Which could free lots of space, but the allocated limit can't be released immediately) |
| Comment by Shuichi Ihara (Inactive) [ 08/May/14 ] |
|
Hi Niu, $ ls -l total 1631392 -rw-r--r-- 1 uits0006 gits 82 Nov 19 09:14 cdbullet.gif drwxr-xr-x 2 uits0006 gits 4096 Nov 19 09:14 css drwxr-xr-x 2 uits0006 gits 4096 Nov 19 09:14 images -rw-r--r-- 1 uits0006 gits 578119 Nov 19 01:44 InstallationGuide_8.06.007_01.html -rw-r--r-- 1 uits0006 gits 1862473 Nov 19 09:16 LicenseAdministration.pdf -rw-r--r-- 1 uits0006 gits 515515 Nov 19 09:14 ReleaseNotes_8.06.007_01.html drwxr-xr-x 2 uits0006 gits 4096 Nov 19 09:13 scripts -rwxr-xr-x 1 uits0006 gits 1667566384 Nov 19 09:09 STAR-CCM+8.06.007_01_linux-x86_64-2.3.4_gnu4.6-r8.bin There are more than 8 files at least(we checked he has 121 files for total), but "lfs quota" says there is only one file for this user. $ lfs quota /home2
Disk quotas for user uits0006 (uid 4879):
Filesystem kbytes quota limit grace files quota limit grace
/home2 6447520 10485760 10485760 - 1 0 0 -
|
| Comment by Li Xi (Inactive) [ 08/May/14 ] |
|
Hi Yawei, Thanks for you explaination. I have a doubt here. I though the grace time is triggered if and only if the actually used space exceeds the softlimit, but it seems that I was wrong? The grace time is triggered iff the allocated(or can I say 'granted') limits exceeds the softlimit, right? So, that means, there is some difference between allocated limits and actually uses space? If so, then what is the difference? Thanks, |
| Comment by Niu Yawei (Inactive) [ 08/May/14 ] |
|
Hi, Ihara Hi, LiXi Did you just deleted some large files just before run the 'lfs quota'? Can this be reproduced? |
| Comment by Shuichi Ihara (Inactive) [ 08/May/14 ] |
|
Niu, |
| Comment by Niu Yawei (Inactive) [ 09/May/14 ] |
Did you ever disable/enable quota feature for mdt device? I suggest you upgrade the e2fsprogs and disable/enable quota feature for mdt device to see if the problem can be resolved. |
| Comment by Shuichi Ihara (Inactive) [ 13/May/14 ] |
|
Hi Niu, We didn't disable/enable quota yet since we needed to umount/mount MDT.We will be going to try this disable/enable, but before that, we really want to makre quota corraption happens even we didn't run any e2fsck or writing data with tool in old e2fsprogs? |
| Comment by Niu Yawei (Inactive) [ 14/May/14 ] |
|
If tune2fs & e2fsck wasn't used on lustre device, the quota accounting file should be complete. Could you try to create new files to see if it'll be accounted? |
| Comment by Shuichi Ihara (Inactive) [ 14/May/14 ] |
|
New files are accountted, but total accountted number of files are incorrect. |
| Comment by Shuichi Ihara (Inactive) [ 14/May/14 ] |
|
And, another problem, as far as we checked "lfs quota -v", total accounted file size is close to sum of each of quota slave accounting size. but sum of quota limit in quota slave is completely wrong of cluster wide quota lmit. |
| Comment by Niu Yawei (Inactive) [ 14/May/14 ] |
Sum of allocated limit should be less than hardlimit of the system, because there could be some limits not allocated to slaves yet. |
| Comment by John Fuchs-Chesney (Inactive) [ 10/Jun/14 ] |
|
Shuichi-san, |