Details
-
Story
-
Resolution: Unresolved
-
Normal
-
None
-
None
-
False
-
-
False
Description
As part of our transition to KFNBC and to pass performance tests, we increased the number of dashboard pods to 5. After several performance enhancements, the load on dashboard usage has gone down significantly. I propose we drop the numer of pods back down to 3, which should be a very small 1 line change. It's more important now that we're adding more controllers for serving.
Here are the results from kpouget2 's test with 3 dashboard pods quoted from slack:
Test configuration: 300 users starting with a delay of 2 seconds RHODS 1.18.0-8 running on OCP v4.10.15 Total of 65 nodes in the cluster 3 m6i.2xlarge master nodes 2 m6a.xlarge infra nodes 50 m6i.2xlarge rhods_compute nodes 10 m6i.2xlarge test_pods_only nodes Test pods running on another cluster 300/300 users succeeded looks all good :) (edited)
Attachments
Issue Links
- relates to
-
RHODS-6447 RHODS Dashboard oauth-proxy crashing with 1000 users
- Closed