Bug #1297
closed
write a manypools teuthology task
Added by Anonymous almost 13 years ago.
Updated over 12 years ago.
Description
Write a teuthology task that creates lots of pools:
tasks:
- manypools:
num_of_pools: NUM
- radosbench:
The task would create NUM pools, each with a single dummy object (to enforce pools to exist on PGs). Then the yaml can run radosbench as the next task.
Nevermind the "track_memory_use" part, that can be done with collectd, and collectd buys us much much more.
- Category set to OSD
- Target version set to v0.33
- Translation missing: en.field_position set to 23
- Translation missing: en.field_position deleted (
27)
- Translation missing: en.field_position set to 15
- Assignee set to Greg Farnum
- Status changed from New to 7
Running final tests now, but this appears to be working properly. Which is odd since it seemed broken before, but maybe I misread the information I was getting. Trying it on a 10k pool create now, anyway, so that ought to turn up any issues.
Ah, nope, it hung again. Without debugging enabled, naturally...
- Target version changed from v0.33 to v0.34
- Translation missing: en.field_story_points set to 3
- Translation missing: en.field_position deleted (
44)
- Translation missing: en.field_position set to 43
- Subject changed from qa: find out max pools per osd (without running out of RAM, without degrading performance) to write a manypools teuthology task
- Status changed from 7 to Resolved
So each pool is actually 8 PGs each by default, so the task just writes 16 1-byte objects to each pool using rados bench.
Also available in: Atom
PDF