-
Bug
-
Resolution: Unresolved
-
Medium
-
None
-
None
-
None
-
3
-
9223372036854775807
This issue was created by maloo for Patrickbot <patrick-bot@mulberrytree.us>
This issue relates to the following test suite run: https://testing.whamcloud.com/test_sets/d0343643-db48-47d1-a022-138149503b76
Test session details:
Client
Lustre: 2.17.61.122
Branch: master
Build: https://build.whamcloud.com/job/lustre-reviews/136811
Kernel: 6.12.0-211.68.1.el10_2.x86_64
Server
Lustre: 2.17.61.122
Branch: master
Build: https://build.whamcloud.com/job/lustre-reviews/136811
Kernel: 6.12.0-211.68.1.el10_2_lustre.x86_64
sanity-pcc test_1a and test_1b fail with "Failed to start copytool agt1 on '<agent node>'" when the suite runs after sanity-sec in the same session.
Cause: sanity-sec test_300 (LU-19921, "test user_principal mount option parsing") unmounts /mnt/lustre on client1 (clients_arr[0]), remounts it with user_principal, unmounts it again, then tries a mount that is expected to fail. It never remounts the normal /mnt/lustre on client1. check_and_cleanup_lustre does not restore it. The next suite, sanity-pcc, uses a remote client as the copytool agent. On that node only /mnt/lustre2 is mounted, so lhsmtool_posix fails:
lhsmtool_posix: '252/2' dev not on a mounted Lustre filesystem
lhsmtool_posix: cannot find a Lustre filesystem mounted at '/mnt/lustre': No such device (19)
The pdsh "ssh exited with exit code 19" is that ENODEV exit status. The agent node's `mount` output in the test log lists only /mnt/lustre2 of type lustre, and its dmesg shows the "Unmounted lustre-client" at the end of sanity-sec test 300, with no later mount of /mnt/lustre.
Seen with the same copytool log on unrelated changes and branches: lustre-reviews change 70957 ps5 (onyx-159vm166, session 91fcb068-7334-419c-a9b4-54245fd9ccfb), lustre-master 2026-11-16 (onyx-156vm119, session 58ed2df1-ffd1-4b66-bcfc-a8f0b3f1984e) and 2026-11-11 (onyx-157vm78), lustre-master-next 2026-11-09 (trevis-154vm185). Also seen on change 68619 ps22 (session 8bbfce12-297e-43f4-a5c5-e334e301a88b, trevis-152vm168), which does not touch this code. Rainshadow had no ticket for the error signature (8 failures in 730 days, none linked).
Suggested fix: at the end of sanity-sec test_300 (or via stack_trap), remount client1 on $MOUNT with the original options, so later suites find /mnt/lustre mounted on every client.
Raised by Patch Watcher run pw-engineer-68619-ps22-eb53f06f45e7 on https://review.whamcloud.com/68619