-
Notifications
You must be signed in to change notification settings - Fork 592
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
Leadership pinning: implementation #23691
Conversation
Use revision id of the topic creation command, but introduce a distinct type.
ad58898
to
ea5d50d
Compare
ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/56084#01927120-f364-46d1-b7c1-7498ca75a083 |
This way several constraints can use it, not just the topic-aware distribution constraint.
This is a struct that will hold everything needed for the leadership pinning constraint.
No functional changes.
Previously, there were two problems with muted groups: 1) If the transfer was not successful, we didn't mute 2) If the transfer was successful, the group remained muted for a long time, preventing reaching optimum. Make the balancer more aggressive by removing the group from the muted set if we've got a leadership notification after a successful transfer.
Usually this means that the health report hasn't yet been updated, it doesn't make sense to mute those groups for long.
Cleanup the execution model for balancing iterations: 1) Allow the balancing fiber to run to completion (with small mandatory intervals between iterations). 2) Don't ignore timer if it fired while the fiber was active. 3) If we are not throttled or activating after acquiring controller leadership, allow to schedule timer sooner (useful for notifications).
For now, just trigger the balancer soon after topic creation.
ea5d50d
to
a91893a
Compare
non flaky failures in https://buildkite.com/redpanda/redpanda/builds/56115#019271fd-bb8c-4ad1-b1ce-b7b2d3ad0ad9:
non flaky failures in https://buildkite.com/redpanda/redpanda/builds/56152#019272b6-d818-4fd6-b67c-2038e07b2821:
non flaky failures in https://buildkite.com/redpanda/redpanda/builds/56179#01927362-4df1-449b-ae90-49619d95286b:
|
Retry command for Build#56115please wait until all jobs are finished before running the slash command
|
a91893a
to
26dbd35
Compare
Retry command for Build#56152please wait until all jobs are finished before running the slash command
|
/ci-repeat 1 |
Retry command for Build#56179please wait until all jobs are finished before running the slash command
|
26dbd35
to
8bd5e66
Compare
/ci-repeat 1 |
Ok looks like the test is stable now |
Jira ref: https://redpandadata.atlassian.net/browse/CORE-7022
TODO: add it to the list of enterprise features.
Backports Required
Release Notes
Features
redpanda.leaders.preference
topic config property ordefault_leaders_preference
cluster config property.