-
Bug
-
Resolution: Duplicate
-
Medium
-
None
-
None
-
None
-
3
-
9223372036854775807
[Claude] With 12 MDTs, the sanity-hsm copytool cannot open files by FID: llapi_hsm_action_begin() on /mnt/lustre2/.lustre/fid/<FID> returns ENOENT for files that exist, so HSM requests never reach STARTED and test_251 and test_254b fail.
sanity-hsm test_251: @@@@@@ FAIL: request on 0x20000040d:0xb5:0x0 is not STARTED on mds1 sanity-hsm test_254b: @@@@@@ FAIL: Expected 52 (!= 0) active archive requests
Copytool log (lhsmtool_posix on onyx-155vm20, mounted on /mnt/lustre2), test_251:
[0x20000040d:0xb5:0x0] action ARCHIVE reclen 72, cookie=0x6abb40ff processing file 'd251.sanity-hsm/f251.sanity-hsm' llapi_hsm_action_begin() on '/mnt/lustre2/.lustre/fid/0x20000040d:0xb5:0x0' failed: No such file or directory (2) Action completed, notifying coordinator cookie=0x6abb40ff, FID=[0x20000040d:0xb5:0x0], hp_flags=0 err=2
test_254b shows the same ENOENT for every archive request (e.g. [0x20000040d:0xb7:0x0]). The files were created on /mnt/lustre on the other client in a directory created with mkdir_on_mdt0, and still exist.
Configuration: 2 MDS nodes, MDSCOUNT=12, from the custom test session for https://review.whamcloud.com/69303 (LU-14058 >10 MDT script fixes), which removes the old MDSCOUNT limit from sanity-hsm.sh. sanity-hsm has not previously been run with more than 10 MDTs.
Maloo auto-linked LU-16932 by error signature, but the cause here is the ENOENT FID lookup on the copytool mount, not a timing issue. Possibly the same root cause as the sanity-hsm test_500 failure with 12 MDTs (stat of a just-created directory returns ENOENT), see linked ticket.
sanity-hsm test_251 and test_254b are excluded with MDSCOUNT > 9 by LU-14058 until this is fixed.