issue:2287: find maximum number of informers that doesn't impact latency. #2306
Conversation
k8s-ci-robot
added
cncf-cla: yes
|
/assign @wojtek-t |
It was more like 6-7 seconds looking into perf-dash. /lgtm |
k8s-ci-robot
added
the
lgtm
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: p0lyn0mial, wojtek-t The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
the
approved
| @@ -22,5 +22,5 @@ spec: | |||
| cpu: "6" | |||
| command: [ "watch-list" ] | |||
| # TODO(p0lyn0mial): bring back 16 informers | |||
| args: [ "--alsologtostderr=true", "--v=4", "--timeout={{.Duration}}", "--count=8", "--namespace=watch-list-1", "--enableWatchListFeature={{.EnableWatchListFeature}}"] | |||
| args: [ "--alsologtostderr=true", "--v=4", "--timeout={{.Duration}}", "--count=6", "--namespace=watch-list-1", "--enableWatchListFeature={{.EnableWatchListFeature}}"] | |||
There was a problem hiding this comment.
nit, we should expand the comment to explain this magic number, is going to be hard to trace back from the git logs
There was a problem hiding this comment.
This is just temporary to gather some data. I am going to bring back the previous value once we finish with the experiment.
What type of PR is this?
After #2305 the latency increased to
30s.With #2303 it was around
1s.Between
1sand30sthe CPU usage didn't significantly increase.Everything indicates that the egress is being throttled.
According to my calculations we should be below the egress limit (2 GB/s) with
6or5informers.This PR set the value of the informers to
6.Once we find our sweet spot I'm going to add a metric that will measure the latency for streaming.
Then we update the value of informers to initial value of
16and compare the latencies one more time.xref: #2287
What this PR does / why we need it:
Which issue(s) this PR fixes:
Fixes #
Special notes for your reviewer: