[LU-9365] Sync data before get df data Created: 19/Apr/17  Updated: 02/Feb/18  Resolved: 02/Feb/18

Status: Closed
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Bug Priority: Minor
Reporter: Yang Sheng Assignee: Yang Sheng
Resolution: Not a Bug Votes: 0
Labels: None

Issue Links:
Related
is related to LU-5788 recovery-double-scale test_pairwise_f... Resolved
Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   
== sanity test 311: disable OSP precreate, and unlink should destroy objs ============================ 22:16:56 (1481926616)
total: 1000 creates in 3.08 seconds: 325.07 creates/second
 - unlinked 0 (time 1481926620 ; total 0 ; last 0)
total: 1000 unlinks in 2 seconds: 500.000000 unlinks/second
waited 120 sec, old Iused 5991, new Iused 9110
 sanity test_311: @@@@@@ FAIL: objs not destroyed after unlink 
  Trace dump:
  = /usr/lib64/lustre/tests/test-framework.sh:4970:error()
  = /usr/lib64/lustre/tests/sanity.sh:14395:test_311()
  = /usr/lib64/lustre/tests/test-framework.sh:5230:run_one()
  = /usr/lib64/lustre/tests/test-framework.sh:5268:run_one_logged()
  = /usr/lib64/lustre/tests/test-framework.sh:5072:run_test()
  = /usr/lib64/lustre/tests/sanity.sh:14397:main()
Dumping lctl log to /tmp/test_logs/1481926605/sanity.test_311.*.1481926744.log

After investigating i think statfs doesn't update in timely manner. So first 'lfs df' may not reflect real filesystem statistic. So we should consider doing a sync action while 'lfs df' in first place. Will push a patch for it.



 Comments   
Comment by Gerrit Updater [ 20/Apr/17 ]

Yang Sheng (yang.sheng@intel.com) uploaded a new patch: https://review.whamcloud.com/26762
Subject: LU-9365 tests: sync data before get df data
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 1519303eced864f1bf9726bc34ab553af30433ce

Comment by Andreas Dilger [ 25/Apr/17 ]

I don't think "sync" will affect the inode counts on the ldiskfs MDS, since they are not cached by the clients. All file creates are done immediately on the MDS so "lfs df -i" should always be accurate.

Generated at Sat Feb 10 02:25:33 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.