Fix CI instance variable cache misses
Previously the CI variable instance cache always attempted to read Redis for a timestamp to determine whether the value was up-to-date. The problem was that in a system with multiple processes reading from the cache, the timestamp would constantly be advancing, making it appear as though the cache always were stale. Not only did this incur one Redis call per lookup, but also this would result in frequent cache misses. Since we mostly just use this cache data set in the creation of pipelines, we can likely live with a 30-second delay between the time when the variable is set and when it is live. For the `gitlab-org/gitlab` project, this should reduce the overall pipeline creation time by around 1.5 seconds (11% of the time), not to mention reduce the network and memory overhead. Relates to https://gitlab.com/gitlab-org/gitlab/-/issues/322386 Changelog: fixed
Showing
Please register or sign in to comment