Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

cluster manager: dynamically inserted clusters should be warming at first #12670

Closed
Shikugawa opened this issue Aug 16, 2020 · 4 comments · Fixed by #12783
Closed

cluster manager: dynamically inserted clusters should be warming at first #12670

Shikugawa opened this issue Aug 16, 2020 · 4 comments · Fixed by #12783

Comments

@Shikugawa
Copy link
Member

This is related issue of #11120.
I think that the startup behavior of clusters which are triggered by addOrUpdate() is strange.
Facing problem is that dynamically inserted cluster which depends on SDS is always active even though depending TLS certificate is not initialized because some of issue occurred, like there is no cert info on DiscoveryResponse.
I think that this problem is caused by current behavior of ClusterManager. In current implementation, clusters inserted dynamically before all of init clusters haven't initialized will be marked immediately active. As a result, there is no indicator that inserted cluster is active or warming.
In this situation, dynamically inserted cluster (calls cluster_1) should walk state change like,

  1. Stats is indicating that cluster_1 is warming when addOrUpdate() is called before all of init clusters are not initialized.
  2. While there is no TLS certificate, SDS subscription belongs to cluster_1's transport socket factory is send DiscoveryRequest repeatedly until it reaches to initial_fetch_timeout, and stats is indicating warming. When it retrieved TLS certificate, the state of cluster_1 should change from warming to active.
  3. After it reaches to initial_fetch_timeout and there is no certificate, stats should indicate cluster_1 is warming eternally.

It may be wrong understanding about cluster manager. But I think that we need to consider the initializing cluster's state change to resolve this problem.

cc. @howardjohn

@mattklein123
Copy link
Member

I think this is a duplicate of #11120 or a more general solution to it? Can we track this issue in one place?

@Shikugawa
Copy link
Member Author

Shikugawa commented Aug 17, 2020

@mattklein123 I think that this is a part of #11120. So not duplicate. More general solution is required. This issue should be focused on Cluster Management. And #11120 should be focused on SDS. In my thought, we should track the changes of behavior of dynamically inserted cluster in this issue.

@stale
Copy link

stale bot commented Sep 20, 2020

This issue has been automatically marked as stale because it has not had activity in the last 30 days. It will be closed in the next 7 days unless it is tagged "help wanted" or other activity occurs. Thank you for your contributions.

@stale stale bot added the stale stalebot believes this issue/PR has not been touched recently label Sep 20, 2020
@Shikugawa
Copy link
Member Author

/nostale

@stale stale bot removed the stale stalebot believes this issue/PR has not been touched recently label Sep 21, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
2 participants