[LU-13576] sanity-quota became much slower Created: 18/May/20  Updated: 30/Apr/21  Resolved: 30/Apr/21

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

Type: Bug Priority: Minor
Reporter: Alex Zhuravlev Assignee: WC Triage
Resolution: Not a Bug Votes: 0
Labels: None
Environment:

ldiskfs


Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   

before very recently sanity-quota took ~2700s locally, after the last landings to the master - >4500s

top slowest subtests,

was:


 38 277
 8 175
 3 171
 7c 154
 4a 151
 22 113
 17 107
 6 100
 59 99
 63 80
 35 77
 7a 70
 34 63
 39 59
 23 59
 

fresh master:

 38 395
 3b 248
 3a 202
 8 184
 7c 170
 4a 161
 3c 144
 1d 132
 1c 131
 17 122
 22 115
 1b 112
 6 111
 63 110
 1f 108


 Comments   
Comment by Alex Zhuravlev [ 18/May/20 ]

it's 13 new tests adding ~1500s, >50% of old sanity-quota.

but in general the majority of existing tests became slower as well.

especailly 38 (which  LU-11023 doesn't touch): 277 to 395 seconds.

 

Comment by Peter Jones [ 18/May/20 ]

Is this because of OST pool quotas adding more tests or something?

Comment by Sergey Cheremencev [ 25/May/20 ]

Probably it may be caused by "trace" debug level added in "LU-11023 quota: quota pools for OSTs".

Comment by Alex Zhuravlev [ 30/Apr/21 ]

that seem to be caused by project quota

Generated at Sat Feb 10 03:02:26 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.