iodepth=1 is hardcoded in read tests
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Woodpecker Charm |
Fix Committed
|
Undecided
|
Unassigned |
Bug Description
The charm allows randwrite tests with iodepth=32 for example. However, the charm always overrides iodepth to 1 in read/randread tests without giving enough information or feedback to users who specified an explicit iodepth.
As the result, read performance is always worse than write, and that confuses people.
https:/
> I don't think this is the right fix for the original issue mentioned. For the record, iodepth > 1 makes sense if block size is 4k although measuring performance of iodepth=1 is always good as a reference.
>
> In any case, just ignoring and overriding iodepth are a bad practice to me.
>
> ```
> [rbd_iodepth32]
> {% if 'read' in action_
> iodepth=1
> ```
>
> More fundamental issue is described in https:/
https:/ /github. com/openstack- charmers/ charm-woodpecke r/pull/ 11