Add soft Pod anti-affinity to IMC dispatchers#4640
Conversation
Codecov Report
@@ Coverage Diff @@
## master #4640 +/- ##
=======================================
Coverage 81.31% 81.31%
=======================================
Files 290 290
Lines 8176 8176
=======================================
Hits 6648 6648
Misses 1128 1128
Partials 400 400 Continue to review full report at Codecov.
|
|
That sounds great, thanks for the great PR description! /assign |
|
I like the idea. Also the picture is amazing. Thanks for the great work. /lgtm |
|
/lgtm |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: antoineco, pierDipi The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
* Add soft Pod anti-affinity to IMC dispatchers (knative#4640) * make RELEASE=v0.19.2 generate-release Signed-off-by: Matthias Wessendorf <mwessend@redhat.com> Co-authored-by: Antoine Cotten <hello@acotten.com>
* Add soft Pod anti-affinity to IMC dispatchers (knative#4640) * make RELEASE=v0.19.2 generate-release Signed-off-by: Matthias Wessendorf <mwessend@redhat.com> Co-authored-by: Antoine Cotten <hello@acotten.com>
Proposed Changes
Dispatchers consume a fair amount of CPU and may starve on CPU time under load, as observed in the graph below (load-test at 24K then 27K events/sec against 5 replicas). Defining requests and limits would be too opinionated because we don't know where Knative is being deployed, but we can at least try to ensure dispatchers are spread across nodes.
Release Note
Docs