volcano-sh
metrics: use milliseconds instead of microseconds as the time unit for scheduling latency
The current time unit cannot accurately reflect the actual scheduling time like below
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="5"} 0
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="10"} 0
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="20"} 0
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="40"} 0
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="80"} 0
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="160"} 0
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="320"} 1
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="640"} 1
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="1280"} 68550
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="2560"} 198253
volcano_action_scheduling_latency_microseconds_bucket{action="allocate",le="+Inf"} 243524
volcano_action_scheduling_latency_microseconds_sum{action="allocate"} 5.505664439029926e+08
Signed-off-by: Liang Zheng [email protected]
Welcome @microyahoo!
It looks like this is your first PR to volcano-sh/volcano.
Thank you, and welcome to Volcano. :smiley:
Shall we mark it deprecated first and then remove it after several releases?
Shall we mark it deprecated first and then remove it after several releases?
hi @lowang-bh, thanks for your quick response. There is also another workaround that doesn't introduce a breaking change; we just need to modify the exponential bucket count to cover a wider range.
What's the problem of using microseconds?
hi @Vacant2333, if microseconds are used, the maximum range of the bucket is 2.56ms. Values greater than 2.56ms will all fall into this bucket, which does not reflect the actual scheduling latency.
What's the problem of using microseconds?
hi @Vacant2333, if microseconds are used, the maximum range of the bucket is 2.56ms. Values greater than 2.56ms will all fall into this bucket, which does not reflect the actual scheduling latency.
I want to know how you @mentioned me; I have no connection to this PR.
What's the problem of using microseconds?
hi @Vacant2333, if microseconds are used, the maximum range of the bucket is 2.56ms. Values greater than 2.56ms will all fall into this bucket, which does not reflect the actual scheduling latency.
I want to know how you @mentioned me; I have no connection to this PR.
I think he's pointing me: )
What's the problem of using microseconds?
hi @Vacant2333, if microseconds are used, the maximum range of the bucket is 2.56ms. Values greater than 2.56ms will all fall into this bucket, which does not reflect the actual scheduling latency.
I want to know how you @mentioned me; I have no connection to this PR.
sorry, my mistake @Vacant2333
[APPROVALNOTIFIER] This PR is APPROVED
This pull-request has been approved by: william-wang
The full list of commands accepted by this bot can be found here.
The pull request process is described here
- ~~pkg/scheduler/OWNERS~~ [william-wang]
Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment