[LU-6532] tests which call wait_delete_completed() are to be skipped on CLIENTONLY mode Created: 28/Apr/15  Updated: 28/Apr/15

Status: Open
Project: Lustre
Component/s: None
Affects Version/s: Lustre 2.7.0
Fix Version/s: None

Type: Bug Priority: Minor
Reporter: Elena Gryaznova Assignee: WC Triage
Resolution: Unresolved Votes: 0
Labels: None

Severity: 3
Epic: test
Rank (Obsolete): 9223372036854775807

 Description   

wait_delete_completed () requires server access :

wait_delete_completed_mds() {
...
       # find MDS with pending deletions
       for node in $(mdts_nodes); do
               changes=$(do_node $node "lctl get_param -n osc.*MDT*.sync_*" 

The tests which call wait_delete_completed () are to be skipped for CLIENTONLY mode, otherwise these tests could fail :

# CLIENTONLY=1 MGSNID=fre0101@tcp OSD_TRACK_DECLARES_LBUG=1 sh runtests

== runtests test 1: All Runtests == 08:13:27 (1430208807)
touching /mnt/lustre at Tue Apr 28 08:13:27 UTC 2015 (@1430208807)
create an empty file /mnt/lustre/hosts.29640
copying /etc/hosts to /mnt/lustre/hosts.29640
comparing /etc/hosts and /mnt/lustre/hosts.29640
renaming /mnt/lustre/hosts.29640 to /mnt/lustre/hosts.29640.ren
copying /etc/hosts to /mnt/lustre/hosts.29640 again
truncating /mnt/lustre/hosts.29640
removing /mnt/lustre/hosts.29640
copying /etc/hosts to /mnt/lustre/hosts.29640.2
truncating /mnt/lustre/hosts.29640.2 to 123 bytes
creating /mnt/lustre/d1.runtests
copying 1000 files from /etc /bin to /mnt/lustre/d1.runtests/etc /bin at Tue Apr 28 08:13:28 UTC 2015
...

disable quota as required
removing /mnt/lustre/d1.runtests
renaming /mnt/lustre/hosts.29640.ren to /mnt/lustre/hosts.29640
truncating /mnt/lustre/hosts.29640
removing /mnt/lustre/hosts.29640
verifying /mnt/lustre/hosts.29640.2 is 123 bytes
running createmany -d /mnt/lustre/d1.runtests/d 100
total: 100 creates in 0.10 seconds: 1005.13 creates/second
running unlinkmany -d /mnt/lustre/d1.runtests/d 100
 - unlinked 0 (time 1430208852 ; total 0 ; last 0)
total: 100 unlinks in 1 seconds: 100.000000 unlinks/second
done
Waiting for local destroys to complete
 runtests test_1: @@@@@@ FAIL: Space not all freed: now 61068kB, was 51536kB. 


 Comments   
Comment by Elena Gryaznova [ 28/Apr/15 ]

It is not reasonable to add "remote_mds_nodsh" check to wait_delete_completed_mds() because tests could run wait_delete_completed_mds() at the end.
There are ~75 tests call wait_delete_completed ():

$ grep wait_delete_completed ./* -r | wc -l
      75

but some of them are already skipped for remote mds with nodsh set.

I do not provide the patch because I'm not sure what is the correct way to fix this defect.
Please advice.
Thanks.

Generated at Sat Feb 10 02:01:02 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.