[LU-4494] sanity-scrub test 4: lustre-MDT0001: too many transaction credits (309 > 256)

• Type: Bug
• Resolution: Duplicate
• Priority: Minor
• Affects Version/s: Lustre 2.6.0
• Labels: dne
• Environment: Lustre branch: master; MDSCOUNT=4
• Severity: 3
• Assignee: WC Triage
• Reporter: Jian Yu

Description

While testing patch http://review.whamcloud.com/8523 on master branch with MDSCOUNT=4, the following issues occurred on one of the MDTs during create:

Lustre: DEBUG MARKER: mkdir -p /mnt/mds4; mount -t lustre -o user_xattr,acl                                /dev/lvm-Role_MDS/P4 /mnt/mds4^M
LDISKFS-fs (dm-3): mounted filesystem with ordered data mode. quota=on. Opts: ^M
LDISKFS-fs (dm-3): mounted filesystem with ordered data mode. quota=on. Opts: ^M
Lustre: DEBUG MARKER: PATH=/usr/lib64/lustre/tests:/usr/lib/lustre/tests:/usr/lib64/lustre/tests:/opt/iozone/bin:/opt/iozone/bin:/usr/lib64/lustre/tests/mpi:/usr/lib64/lustre/tests/racer:/usr/lib64/lustre/../lustre-iokit/sgpdd-survey:/usr/lib64/lustre/tests:/usr/lib64/lustre/u^M
Lustre: DEBUG MARKER: e2label /dev/lvm-Role_MDS/P4 2>/dev/null^M
Lustre: DEBUG MARKER: /usr/sbin/lctl mark Using TIMEOUT=20^M
Lustre: DEBUG MARKER: Using TIMEOUT=20^M
Lustre: cli-ctl-lustre-MDT0001: Allocated super-sequence [0x0000000440000400-0x0000000480000400):1:mdt]^M
Lustre: 32030:0:(osd_handler.c:872:osd_trans_start()) lustre-MDT0001: too many transaction credits (309 > 256)
Lustre: 32030:0:(osd_handler.c:879:osd_trans_start())   create: 2/50, delete: 0/0, destroy: 0/0
Lustre: 32030:0:(osd_handler.c:884:osd_trans_start())   attr_set: 2/2, xattr_set: 4/43
Lustre: 32030:0:(osd_handler.c:891:osd_trans_start())   write: 11/154, punch: 2/8, quota 2/2
Lustre: 32030:0:(osd_handler.c:896:osd_trans_start())   insert: 3/50, delete: 0/0
Lustre: 32030:0:(osd_handler.c:901:osd_trans_start())   ref_add: 0/0, ref_del: 0/0
Pid: 32030, comm: mdt00_002

Call Trace:
 [<ffffffffa0903895>] libcfs_debug_dumpstack+0x55/0x80 [libcfs]
 [<ffffffffa0a95a43>] osd_trans_start+0x653/0x670 [osd_ldiskfs]
 [<ffffffffa0dfc579>] lod_trans_start+0x1b9/0x250 [lod]
 [<ffffffffa0e72317>] mdd_trans_start+0x17/0x20 [mdd]
 [<ffffffffa0e65dd9>] mdd_create+0x929/0x17a0 [mdd]
 [<ffffffffa0d52fa8>] mdo_create+0x18/0x50 [mdt]
 [<ffffffffa0d5d037>] mdt_reint_open+0x12a7/0x2120 [mdt]
 [<ffffffffa0d47a91>] mdt_reint_rec+0x41/0xe0 [mdt]
 [<ffffffffa0d2ddb3>] mdt_reint_internal+0x4c3/0x780 [mdt]
 [<ffffffffa0d2e265>] mdt_intent_reint+0x1f5/0x520 [mdt]
 [<ffffffffa0d2c989>] mdt_intent_policy+0x499/0xca0 [mdt]
 [<ffffffffa0527509>] ldlm_lock_enqueue+0x359/0x920 [ptlrpc]
 [<ffffffffa0550edf>] ldlm_handle_enqueue0+0x4ef/0x10b0 [ptlrpc]
 [<ffffffffa05cb492>] tgt_enqueue+0x62/0x1d0 [ptlrpc]
 [<ffffffffa05cdfea>] tgt_handle_request0+0x2ea/0x1490 [ptlrpc]
 [<ffffffffa05cf5ca>] tgt_request_handle+0x43a/0x980 [ptlrpc]
 [<ffffffffa0582725>] ptlrpc_main+0xd25/0x1970 [ptlrpc]

And then the nodes were re-provisioned by autotest.

Maloo report: https://maloo.whamcloud.com/test_sets/12c1c02a-7e6f-11e3-925a-52540035b04c

Issue Links

• is related to: LU-4611 too many transaction credits (32279 > 25600)

Comments

[Andreas Dilger]

Closing as a duplicate of LU-4611, which shows the problem much more significantly with a large number of stripes.

[Oleg Drokin]

Well, it almost looks like one of the filesystems was formatted with 4M journal size (256 blocks max transactions).
So then when you try to do a create it fails.

Could it just be that mdscount=4 overwhelms the journal size by doing a larger transaction than in normal case for some sort of a special purpose small filesystem?
If so, this woul likely be visible on any test run run in such conditions and we probably can do nothing about it other than changing the fs journal size used for this test