[LU-8851] nodemap: add flags to limit mapping to UID or GID only

• Status: Resolved
• Project: Lustre
• Fix Version/s: Lustre 2.10.0
• Type: Improvement
• Priority: Minor
• Reporter: Kit Westneat
• Assignee: Kit Westneat
• Resolution: Fixed

Description

There could be cases where different systems have the same UIDs but different GIDs, or vice versa, and therefore need mapping only for UIDs or GIDs. There should be a way to enable only UID mapping or only GID mapping.

Issue Links

• is related to: LU-13172 nodemap: a squashed primary GID allow...
• is related to: LU-3291 IU UID/GID Mapping Feature
• is related to: LU-9671 Group quota not enforced on clients w...

Comments

[Gerrit Updater]

Kit Westneat ([e-mail]) uploaded a new patch: http://review.whamcloud.com/23853
Subject: LU-8851 nodemap: add uid/gid only flags to control mapping
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 0ae6cd826a556c30d5d5ae4f68b933e398ed6935

[Stephane Thiell]

So far this patch works fine with the 2.9 community release.

I have used it quite a lot and only a single time I noticed an error message when using nodemap_add_idmap but it did work as expected in the end:

[3785311.191429] LustreError: 117363:0:(mgs_handler.c:779:mgs_iocontrol_nodemap()) MGS: OBD_IOC_NODEMAP command CE046 for xstream: rc = -2

Don't worry, I will report any other suspicious behavior.

I guess the patch has several green lights from reviewers. Would Intel be ok to provide this patch in 2.9 FE?

Thanks!

[Stephane Thiell]

Hi Kit,

We have been using this patch for some time now, it is a critical patch to us and we're very happy to have it. We have noticed that lfs quota is broken from the GID-only clients with nodemap enabled.

Example canonical GID 3199 on Oak mapped to 2709 on XStream:

[root@oak-rbh01 ~]# lfs quota -gh oak_ruthm /oak
Disk quotas for group oak_ruthm (gid 3199):
     Filesystem    used   quota   limit   grace   files   quota   limit   grace
           /oak  4.028T  9.313T     10T       -  115749  1450000 1500000       -

[sthiell@xstream-ln01 ~]$ lfs quota -hg srcc /oak
Disk quotas for group srcc (gid 2709):
     Filesystem    used   quota   limit   grace   files   quota   limit   grace
           /oak  4.028T      0k      0k       -       0       0       0       -

We have both group disk and inodes quota enabled but only the disk usage is reported, weird heh?

We're seeing the same behaviour on both Lustre 2.5 (Cray C3) and IEEL 3.0 (Lustre 2.7 based) clients. Intel, this is also tracked in DELL-291.

I also see that a quota test failed. I checked your patch, but I didn't find anything related to quota. Kit, do you have an idea of what could be wrong here?

Thanks much!
Stephane

[Kit Westneat]

Hi Stephane,

I'm glad the patch is working out! While I was digging into this, I found a typo in the nodemap quota code where if you check group quota, it actually maps the UID. I created LU-9258 and uploaded:
https://review.whamcloud.com/26209

Can you check out this patch and see if it fixes your issues?

Thanks,
Kit

[Stephane Thiell]

Hi Kit,

Good catch!!! Will do ASAP and let you know how it goes. Thanks so much!

All the best,
Stephane

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/23853/
Subject: LU-8851 nodemap: add uid/gid only flags to control mapping
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: f7815c80d95b20c97317fbd7010013dca009121e

[Peter Jones]

Landed for 2.10