[LU-4647] Add idmapping functions for nodemap

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.6.0
• Fix Version/s: Lustre 2.8.0
• Type: Technical task
• Priority: Critical
• Reporter: Joshua Walgenbach (Inactive)
• Assignee: Nathaniel Clark
• Resolution: Fixed
• Labels: HB, patch

Issue Links

• is related to: LU-5963 wrong hashing in LU-4647
• is related to: LU-1882 conf-sanity test_45: umount2: Device ...
• is related to: LUDOC-265 Nodemap lctl functions need to be doc...

Comments

[Andreas Dilger]

http://review.whamcloud.com/9299

[Jodi Levi (Inactive)]

http://review.whamcloud.com/#/c/9529/

[Andreas Dilger]

I noticed http://review.whamcloud.com/7936 coming down the pipe as well, and while the client-side changes obviously don't affect this feature, it will need some idmap handling on the server side.

[Kit Westneat]

Added tests here:
http://review.whamcloud.com/10406

[Kit Westneat]

I was doing some testing today and ran into a weird issue. Basically runas doesn't seem to work unless there is a supplemental group (it doesn't matter what it is):

[root@c1 mnt]# /root/lustre-release/lustre/tests/runas -u 503  touch `pwd`/tmp_dir/file_test
running as uid/gid/euid/egid 503/503/503/503, groups:
 [touch] [/mnt/tmp_dir/file_test]
touch: cannot touch `/mnt/tmp_dir/file_test': Permission denied
[root@c1 mnt]# /root/lustre-release/lustre/tests/runas -u 503 -G99999 touch `pwd`/tmp_dir/file_test
adding supp group 99999
running as uid/gid/euid/egid 503/503/503/503, groups: 99999
 [touch] [/mnt/tmp_dir/file_test]

I wanted to check in to see if anyone else can reproduce it, or if it's something about my build/system. Also if anyone has any theories as to why it's happening.

Thanks,
Kit

[Andreas Dilger]

This should probably be a separate bug.

[Cliff White (Inactive)]

Kit, can you open your issue as a new bug, please?

[Cliff White (Inactive)]

Kit, your patch http://review.whamcloud.com/#/c/10406/ is failing testing, can you address the issues?

[Kit Westneat]

Hi Cliff, I'm not sure what's going on. It seems like it's having a problem creating a file as root. Do these nodes run with root squash on or anything? Is there a way to get interactive access to the test or is it batch only? I can't reproduce the issue on my test machine.

[Kit Westneat]

I think I figured it out - the earlier tests were enabling nodemap and the new tests expect it to be disabled at start. I uploaded a fix for it.

There still seems to be a problem with the tests, but it's failing on other non-nodemap tests as well. They are all getting timeout errors. Is it possible that there is a problem with the test cluster?

[Kit Westneat]

Looking at it again, it looks like the OSS kernel panic'd:
22:04:24:IP: [<ffffffffa08882e5>] tgt_request_preprocess+0x705/0x1470 [ptlrpc]

I dug into it some, and it looks like the clients aren't being reassigned a nodemap after the nodemap_destroy, so they are just pointing to NULL nodemaps. Josh is working on an updated patch and will resubmit it.

[Cliff White (Inactive)]

Any updates on this ? The last patch also failed testing.

[Kit Westneat]

Found a new bug while working on 10406:
http://review.whamcloud.com/#/c/12575/

[Alexey Lyashkov]

I hope Intel will block to close that ticket until oxygen comments will don't added to the code with good debug options.
It's very strange to land code without ANY debug in code. Why Intel allow to do it? did Intel do any testing? recovery-small failed to stop any service with export reference leak with simple test SLOW=yes sh recovery-small.sh.

[Cliff White (Inactive)]

Alex, this is in reference to http://review.whamcloud.com/#/c/12575/ ?

[Alexey Lyashkov]

Cliff,

I don't know a patch pointed you. my patch in LU-5963 very simple.

@@ -50,7 +52,12 @@ void nm_member_putref(struct obd_export *exp)
 static __u32 nm_member_hashfn(cfs_hash_t *hash_body,
                           const void *key, unsigned mask)
 {
-       return cfs_hash_djb2_hash(key, strlen(key), mask);
+       unsigned long hash;
+//     return cfs_hash_djb2_hash(key, strlen(key), mask);
+       hash = hash_long((unsigned long)key, hash_body->hs_bkt_bits) & mask;
+
+       CDEBUG(D_INFO, "hashing %p -> %lu\n", key, hash);
+       return hash;
 }

but some debug need added anyway.

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/12575/
Subject: LU-4647 nodemap: fix problem with node reclassification
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: ec6aae2e3aa4174e3c8fdd8e75a9a4186e3abdc4

[Jodi Levi (Inactive)]

Patch landed to Master.

[Andreas Dilger]

I think the recently landed patch is not addressing the issue that Alexey pointed out.

[Andreas Dilger]

Sorry, I see that the hash function patch landed via http://review.whamcloud.com/12881 from LU-5963.

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/10406/
Subject: LU-4647 nodemap: add tests to sanity-sec for nodemap mapping
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 50f95dd6e5159981f6b950d2fab674955442603b

[Gerrit Updater]

Andreas Dilger ([e-mail]) uploaded a new patch: http://review.whamcloud.com/13028
Subject: LU-4647 tests: enable sanity-sec for DNE mode
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 44d7be42028c31c49c9929053ea77adae3abde8e

[Gerrit Updater]

Kit Westneat ([e-mail]) uploaded a new patch: http://review.whamcloud.com/13343
Subject: LU-4647 tests: modify sanity-sec to test multiple MDSes
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: e2de239ed18bd9a346e5c4deb1e2762f6d6b07d0

[Jodi Levi (Inactive)]

Reopening based on comments and new patches

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/13478/
Subject: LU-4647 lctl: add nodemap man pages to lctl
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: e477f21ed67bfb76aba2b7cede8f62cc0cbd2ff7

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/13343/
Subject: LU-4647 tests: modify sanity-sec to test multiple MDSes
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: e62ce6e1252b33b37130556f31ceb22e8410dfcb

[Gerrit Updater]

Andreas Dilger ([e-mail]) uploaded a new patch: http://review.whamcloud.com/14769
Subject: LU-4647 tests: clean up sanity-sec code style issues
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: d4240b487f7b16ac69f46452192bffdc475d6aae

[Gerrit Updater]

Dmitry Eremin ([e-mail]) uploaded a new patch: http://review.whamcloud.com/14933
Subject: LU-4647 tests: properly cleanup after test failure
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: af43e0b4c60f7e7d1ab7202bfab66785fe6edd52

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/14933/
Subject: LU-4647 tests: properly cleanup after test failure
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: c19992eb6809170acf885f1895fec628982b05ad

[Gerrit Updater]

Andreas Dilger ([e-mail]) merged in patch http://review.whamcloud.com/14769/
Subject: LU-4647 tests: clean up sanity-sec code style issues
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 4fab3b2dfd4225f096e9a38ac14bcf61c41a81d4

[Peter Jones]

Landed for 2.8