Investigate performance in cluster manager pages / management store

[richard-cox]

• During a customer call Neil and myself witness a badly performing dashboard. The Rancher instance had 15+ clusters and lots of third party auth provider users
• When logging in to the dashboard with a user that only had limited permissions on a single cluster (where they could not see as many resources) the performance was greatly improved
• There are lots of investigations and optimisations around general performance with clusters that contain large amount of resources, however we haven't investigated the impact of a large or struggling management store / local cluster
  • Is there a lot of socket churn? What is the socket churn like for systems with lots of auth provider based users?
  • How does a large management store generally affect the cluster side nav (count churn causing redraw?)
  • Should Sean's work pulling out store based processing into a worker node also be applied to the management store?

[nwmac]

@richard-cox Do we still need to do this for next2?

[richard-cox]

Not needed for next2. With https://github.com/rancher/dashboard/pull/8717 there is an option to improve the socket performance, but we should understand if there does resolve these kind of issues or if there's anything additional we need to do

[zube[bot]]

richard-cox said: Some of this will hopefully be addressed via https://github.com/rancher/dashboard/issues/9548