[LU-7275] conf-sanity.sh does not work when run out of build tree Created: 09/Oct/15  Updated: 15/Dec/15  Resolved: 15/Dec/15

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

Type: Bug Priority: Minor
Reporter: Oleg Drokin Assignee: WC Triage
Resolution: Duplicate Votes: 0
Labels: None

Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   

Some time ago conf-sanity was broken It now always fails in test1 as:

modules unloaded.
Resetting fail_loc on all nodes...done.
PASS 0 (30s)

== conf-sanity test 1: start up ost twice (should return errors) == 13:52:43 (1444413163)
start mds service on centos6-6.localnet
Starting mds1:   -o loop /tmp/lustre-mdt1 /mnt/mds1
Start of /tmp/lustre-mdt1 on mds1 failed 1
 conf-sanity test_1: @@@@@@ FAIL: MDS start failed 
  Trace dump:
  = ./../tests/test-framework.sh:4784:error_noexit()
  = ./../tests/test-framework.sh:4815:error()
  = conf-sanity.sh:313:test_1()
  = ./../tests/test-framework.sh:5062:run_one()
  = ./../tests/test-framework.sh:5099:run_one_logged()
  = ./../tests/test-framework.sh:4964:run_test()
  = conf-sanity.sh:321:main()
Dumping lctl log to /tmp/test_logs/1444413099/conf-sanity.test_1.*.1444413172.log
Dumping logs only on local client.
FAIL 1 (9s)

Discussion I had with WangDi:

[7/3/15 2:07:07 PM] wangdi: this because of test script problem,  I believe load_modules is missing in the script
[7/3/15 2:07:41 PM] wangdi: some one changed the script or utils somewhere to make cleanup unload the module
[7/3/15 2:07:57 PM] wangdi: but did not upgrade the test-script, especially conf-sanity
[7/3/15 2:08:15 PM] wangdi: so cont-sanity only works in RPM environment
[7/3/15 2:08:34 PM] wangdi: I did not track which land, but it is a known problem long time ago

And indeed, no lustre modules (other than libcfs) loaded after failure:

[root@centos6-6 tests]# lsmod
Module                  Size  Used by
libcfs                373404  0 
exportfs                4108  0 
dm_log                  9794  2 dm_mirror,dm_region_hash
dm_mod                 99360  2 dm_mirror,dm_log
nfs                   424547  2 
lockd                  73530  1 nfs
fscache                54161  1 nfs
auth_rpcgss            45749  1 nfs
nfs_acl                 2647  1 nfs
sunrpc                265418  15 nfs,lockd,auth_rpcgss,nfs_acl
be2iscsi              105067  0 
bnx2i                  47492  0 
cnic                   56029  1 bnx2i
uio                    10510  1 cnic
cxgb3i                 24459  0 
libcxgbi               55410  1 cxgb3i
ipv6                  352982  46 cnic,libcxgbi
cxgb3                 149584  1 cxgb3i
mdio                    4769  1 cxgb3
libiscsi_tcp           15932  2 cxgb3i,libcxgbi
qla4xxx               275293  0 
iscsi_boot_sysfs        9458  2 be2iscsi,qla4xxx
libiscsi               48212  6 be2iscsi,bnx2i,cxgb3i,libcxgbi,libiscsi_tcp,qla4xxx
scsi_transport_iscsi    99364  5 be2iscsi,bnx2i,libcxgbi,qla4xxx,libiscsi


 Comments   
Comment by Dmitry Eremin (Inactive) [ 12/Oct/15 ]

Usually I do llmount.sh before sanity.sh and it works fine for me.

Comment by Oleg Drokin [ 12/Oct/15 ]

but then there's no unounting.
Also note it's conf-sanity that I report about here.

Comment by Peter Jones [ 15/Dec/15 ]

Di has pushed a fix for this under LU-7555

Generated at Sat Feb 10 02:07:29 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.