Skip to content

Cache namespace first Auto DevOps config

Drew Blessing requested to merge dblessing_group_projects_api_iteration2 into master

What does this MR do and why?

Related to #330140 (closed)

Resolves an N+1 by caching lookup of first_auto_devops_config for 1 day for each group. This configuration is unlikely to change often and is an expensive lookup so it makes sense to cache it.

The cache is proactively expired if any group ancestor of a given group changes the auto_devops_enabled configuration.

Of note, as currently written, this change will store the cache for all namespaces on the same Redis instance/shard. It's required that the cache for a given hierarchy be on the same instance so we can efficiently expire the cache. However, it would be expensive to get the root_ancestor to group the cache based on that.

In review we will explore whether this is acceptable and what other more efficient options are available.

Feature flag

Rollout issue: #353503 (closed)

This MR includes the feature flag namespaces_cache_first_auto_devops_config, which is disabled by default. The flag is included out of an abundance of caution, because we know we have large customers that rely on this endpoint, and because the change is in a high risk area (Namespace model).

What is something goes wrong?

No data loss should occur. At worst, the API will return the wrong information, or perhaps AutoDevops will run or not run if the cache is incorrect. However the risk is not large.

Screenshots or screen recordings

These are strongly recommended to assist reviewers and reduce the time to merge your change.

How to set up and validate locally

Numbered steps to set up and validate the change are strongly suggested.

MR acceptance checklist

This checklist encourages us to confirm any changes have been analyzed to reduce risks in quality, performance, reliability, security, and maintainability.

Edited by Vitali Tatarintev

Merge request reports