[LU-9334] LBUG lu_device_fini()) ASSERTION( atomic_read(&d->ld_ref) == 0 ) failed: Refcount is 1

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.10.0
• Affects Version/s: Lustre 2.10.0
• Environment: Soak stress cluster - Lustre-master build 3554 version=2.9.55_35_gaa32cc5
• Severity: 3
• Assignee: nasf (Inactive)
• Reporter: Cliff White (Inactive)

Description

After installing build 3554, attempting to re-mount the filesystem. On MGS/MDS node:

Apr 13 14:14:42 soak-8 kernel: LDISKFS-fs (dm-5): recovery complete
Apr 13 14:14:42 soak-8 kernel: LDISKFS-fs (dm-5): mounted filesystem with ordered data mode. Opts: user_xattr,user_xattr,errors=remount-ro,no_mbcache,nodelalloc
Apr 13 14:14:44 soak-8 kernel: Lustre: MGS: Connection restored to 7703d69c-e9bc-c7f1-f9af-daf27057ff03 (at 0@lo)
Apr 13 14:14:44 soak-8 kernel: Lustre: soaked-MDT0000: Imperative Recovery not enabled, recovery window 300-900
Apr 13 14:14:44 soak-8 kernel: LustreError: 5015:0:(lfsck_layout.c:6780:lfsck_layout_setup()) soaked-MDT0000-osd: fail to init layout LFSCK component: rc = -20
Apr 13 14:14:44 soak-8 kernel: LustreError: 5015:0:(mdd_device.c:1084:mdd_prepare()) soaked-MDD0000: failed to initialize lfsck: rc = -20
Apr 13 14:14:44 soak-8 kernel: LustreError: 5015:0:(obd_mount_server.c:1840:server_fill_super()) Unable to start targets: -20
Apr 13 14:14:44 soak-8 kernel: Lustre: Failing over soaked-MDT0000
Apr 13 14:14:44 soak-8 kernel: LustreError: 5553:0:(osp_object.c:527:osp_attr_get()) soaked-MDT0001-osp-MDT0000:osp_attr_get update error [0x200000009:0x1:0x0]: rc = -5
Apr 13 14:14:44 soak-8 kernel: LustreError: 5554:0:(lod_sub_object.c:959:lod_sub_prep_llog()) soaked-MDT0000-mdtlov: can't get id from catalogs: rc = -5
Apr 13 14:14:44 soak-8 kernel: LustreError: 5554:0:(lod_dev.c:419:lod_sub_recovery_thread()) soaked-MDT0002-osp-MDT0000 getting update log failed: rc = -5
Apr 13 14:14:44 soak-8 kernel: LustreError: 5553:0:(osp_object.c:527:osp_attr_get()) Skipped 2 previous similar messages
Apr 13 14:14:44 soak-8 kernel: LustreError: 5015:0:(lu_object.c:1224:lu_device_fini()) ASSERTION( atomic_read(&d->ld_ref) == 0 ) failed: Refcount is 1
Apr 13 14:14:44 soak-8 kernel: LustreError: 5015:0:(lu_object.c:1224:lu_device_fini()) LBUG
Apr 13 14:14:44 soak-8 kernel: Pid: 5015, comm: mount.lustre
Apr 13 14:14:44 soak-8 kernel: #012Call Trace:
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0bcb7f3>] libcfs_debug_dumpstack+0x53/0x80 [libcfs]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0bcb861>] lbug_with_loc+0x41/0xb0 [libcfs]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0d0fe48>] lu_device_fini+0xb8/0xc0 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0cf4b12>] ls_device_put+0x82/0x2a0 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0cf4e0d>] local_oid_storage_fini+0xdd/0x210 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa129a308>] mgs_fs_cleanup+0x88/0xa0 [mgs]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa1292b96>] mgs_device_fini+0x196/0x5b0 [mgs]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0cff724>] class_cleanup+0x7f4/0xd80 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0d02364>] class_process_config+0x1f84/0x2c30 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0cf1659>] ? lprocfs_counter_add+0xf9/0x160 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0d030ff>] class_manual_cleanup+0xef/0x810 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0d332e0>] server_put_super+0xb20/0xcd0 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0d3708b>] server_fill_super+0xcdb/0x184c [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0d0f1d8>] lustre_fill_super+0x328/0x950 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0d0eeb0>] ? lustre_fill_super+0x0/0x950 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffff81201b1d>] mount_nodev+0x4d/0xb0
Apr 13 14:14:44 soak-8 kernel: [<ffffffffa0d06c78>] lustre_mount+0x38/0x60 [obdclass]
Apr 13 14:14:44 soak-8 kernel: [<ffffffff812024c9>] mount_fs+0x39/0x1b0
Apr 13 14:14:44 soak-8 kernel: [<ffffffff8121e24f>] vfs_kern_mount+0x5f/0xf0
Apr 13 14:14:44 soak-8 kernel: [<ffffffff812207ae>] do_mount+0x24e/0xaa0
Apr 13 14:14:44 soak-8 kernel: [<ffffffff81185a7e>] ? __get_free_pages+0xe/0x50
Apr 13 14:14:44 soak-8 kernel: [<ffffffff81221096>] SyS_mount+0x96/0xf0
Apr 13 14:14:44 soak-8 kernel: [<ffffffff81696c49>] system_call_fastpath+0x16/0x1b
Apr 13 14:14:45 soak-8 kernel:
Apr 13 14:14:45 soak-8 kernel: Kernel panic - not syncing: LBUG

vmcore-dmesg is attached, crash dump is available on soak-8

Issue Links

• is related to: LU-9415 lfsck_load_one_trace_file() is too noisy
• is related to: LU-8998 Progressive File Layout (PFL)

Comments

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/26703/
Subject: LU-9334 lfsck: object leak in lfsck_load_one_trace_file
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 1b136613440bd81e284a12df97618f92c9729d71

[Gerrit Updater]

Jinshan Xiong ([e-mail]) merged in patch https://review.whamcloud.com/26718/
Subject: LU-9334 lfsck: reset trace file for upgrade case
Project: fs/lustre-release
Branch: pfl
Current Patch Set:
Commit: ba4a83aa9f72d7855ff77a37897619155313b198

[Gerrit Updater]

Andreas Dilger ([e-mail]) uploaded a new patch: https://review.whamcloud.com/26718
Subject: LU-9334 lfsck: reset trace file for upgrade case
Project: fs/lustre-release
Branch: pfl
Current Patch Set: 1
Commit: 740e28ad8fad95080da4b0275562378ef0b51cd3

[Gerrit Updater]

Fan Yong ([e-mail]) uploaded a new patch: https://review.whamcloud.com/26716
Subject: LU-9334 lfsck: reset trace file for upgrade case
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 48672c55bf83bbadec89bbd3406489251d701db0

[Gerrit Updater]

Jinshan Xiong ([e-mail]) merged in patch https://review.whamcloud.com/26715/
Subject: LU-9334 lfsck: object leak in lfsck_load_one_trace_file
Project: fs/lustre-release
Branch: pfl
Current Patch Set:
Commit: e62a57f86027582c8c9b56a7d3d348fb38510f4b

[Gerrit Updater]

Jinshan Xiong ([e-mail]) uploaded a new patch: https://review.whamcloud.com/26715
Subject: LU-9334 lfsck: object leak in lfsck_load_one_trace_file
Project: fs/lustre-release
Branch: pfl
Current Patch Set: 1
Commit: 806549e6ba5e96059cb41d006290bfc70ce58709

[Cliff White (Inactive)]

Tested latest patch

Apr 18 16:26:55 soak-8 kernel: Lustre: MGS: Connection restored to c060ba9e-ea2b-a109-ec49-a8f45fcb1eaf (at 0@lo)
Apr 18 16:26:56 soak-8 kernel: Lustre: soaked-MDT0000: Imperative Recovery not enabled, recovery window 300-900
Apr 18 16:26:56 soak-8 kernel: LustreError: 12003:0:(lfsck_layout.c:6780:lfsck_layout_setup()) soaked-MDT0000-osd: fail to init layout LFSCK component: rc = -20
Apr 18 16:26:56 soak-8 kernel: LustreError: 12003:0:(mdd_device.c:1084:mdd_prepare()) soaked-MDD0000: failed to initialize lfsck: rc = -20
Apr 18 16:26:56 soak-8 kernel: LustreError: 12003:0:(obd_mount_server.c:1840:server_fill_super()) Unable to start targets: -20
Apr 18 16:26:56 soak-8 kernel: Lustre: Failing over soaked-MDT0000
Apr 18 16:26:56 soak-8 kernel: LustreError: 12458:0:(osp_object.c:527:osp_attr_get()) soaked-MDT0001-osp-MDT0000:osp_attr_get update error [0x200000009:0x1:0x0]: rc = -5
Apr 18 16:26:56 soak-8 kernel: LustreError: 12459:0:(lod_sub_object.c:959:lod_sub_prep_llog()) soaked-MDT0000-mdtlov: can't get id from catalogs: rc = -5
Apr 18 16:26:56 soak-8 kernel: LustreError: 12459:0:(lod_dev.c:419:lod_sub_recovery_thread()) soaked-MDT0002-osp-MDT0000 getting update log failed: rc = -5
Apr 18 16:26:56 soak-8 kernel: LustreError: 12458:0:(osp_object.c:527:osp_attr_get()) Skipped 2 previous similar messages
Apr 18 16:27:02 soak-8 kernel: Lustre: 12003:0:(client.c:2113:ptlrpc_expire_one_request()) @@@ Request sent has timed out for slow reply: [sent 1492532816/real 1492532816]  req@ffff8803fc650000 x1565032298054672/t0(0) o251->MGC192.168.1.108@o2ib10@0@lo:26/25 lens 224/224 e 0 to 1 dl 1492532822 ref 2 fl Rpc:XN/0/ffffffff rc 0/-1
Apr 18 16:27:02 soak-8 kernel: Lustre: server umount soaked-MDT0000 complete
Apr 18 16:27:02 soak-8 kernel: LustreError: 12003:0:(obd_mount.c:1502:lustre_fill_super()) Unable to mount  (-20)
Apr 18 16:27:02 soak-8 sshd[11981]: Received disconnect from 192.168.1.135: 11: disconnected by user
Apr 18 16:27:02 soak-8 sshd[11981]: pam_unix(sshd:session): session closed for user root

fails

[Cliff White (Inactive)]

This is on the soak performance cluster - 4 MDS, 4 MDT in a DNE configuration.
Formatted quite some time ago, actually.
MDTs created with older 2.9 build

mkfs.lustre --mgsnode=192.168.1.108@o2ib0:192.168.1.109@o2ib0 --fsname=soaked --mdt --index=1 --servicenode=192.168.1.109@o2ib0:192.168.1.108@o2ib0 --reformat --mkfsoptions='-i 4096' --backfstype=ldiskfs --device-size=2147483648 /dev/disk/by-id/dm-name-360080e50002ff4f00000026d52013098p1
mkfs.lustre --mgsnode=192.168.1.108@o2ib0:192.168.1.109@o2ib0 --fsname=soaked --mdt --index=3 --servicenode=192.168.1.111@o2ib0:192.168.1.110@o2ib0 --reformat --mkfsoptions='-i 4096' --backfstype=ldiskfs --device-size=2147483648 /dev/disk/by-id/dm-name-360080e50001fedb80000015752012949
mkfs.lustre --mgsnode=192.168.1.108@o2ib0:192.168.1.109@o2ib0 --fsname=soaked --mdt --index=2 --servicenode=192.168.1.110@o2ib0:192.168.1.111@o2ib0 --reformat --mkfsoptions='-i 4096' --backfstype=ldiskfs --device-size=2147483648 /dev/disk/by-id/dm-name-360080e50001fedb80000015552012932
mkfs.lustre --mgs --fsname=soaked --mdt --index=0 --servicenode=192.168.1.108@o2ib0:192.168.1.109@o2ib0 --reformat --mkfsoptions='-i 4096' --backfstype=ldiskfs --device-size=2147483648 /dev/disk/by-id/dm-name-360080e50002ff4f00000026952013088p1

To produce the failure, I just mount the MDT. Mount script halts at that point.
Does PFL not work on an existing filesystem? Does the filesystem have to be re-created with PFL code?

[Gerrit Updater]

Fan Yong ([e-mail]) uploaded a new patch: https://review.whamcloud.com/26703
Subject: LU-9334 lfsck: object leak in lfsck_load_one_trace_file
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 6886869d700a7e84f7cc3dc5765096277c84c81e

[nasf (Inactive)]

On the other hand, is it only reproduced on the build 3554? or on other builds also?

[nasf (Inactive)]

Cliff,

Thanks for collecting the logs. Unfortunately, the logs do not contain any information about the failure in the bug description. So I have to try to reproduce it by myself. I have simulated the cases of lfsck_layout_setup() failure and osp_attr_get() failure. But still cannot reproduce it. So would you please to describe your test configuration, including the MDTs count and how to format the MDT devices? And also the steps (commands) to reproduce the issue?

Thanks!