mirror of
https://github.com/hanzoai/tasks.git
synced 2026-10-04 02:04:03 +00:00
Carly de FrondevilleandGitHub
41993fb63f
## What changed? Add an LRU cache that reachability calls before reaching out to visibility. Separate cache for open and closed workflow executions so that there can be separate TTL. TTL configurable via dynamic config. Include cache hit/miss in `reachabilityExitPoint` so that it can be logged and tracked in prometheus, making there be 6 `reachabilityExitPoint` options. ## Why? To reduce load on visibility. ## How did you test it? Tested with unit tests and existing functional tests. Added additional functional test that exercises the reachability cache by doing: 1. Add assignment rule `A` and start workflow with build id `A` 2. Wait for `A` to show up as `running` in visibility db 3. Commit a different build id --> `A` should now only be reachable via visibility query 4. Query `reachability(A)` --> `reachable` by visibility db, populating reachability open WF cache with `A: true` 5. Signal workflow with build id `A` so that it completes 6. Wait for `A` to show up as `closed` in visibility db (cache TTL must be longer than this wait time) 7. Query `reachability(A)` --> `reachable` by reachability cache 8. Wait for `testReachabilityCacheOpenWFsTTL` to pass 9. Query `reachability(A)` --> `closed_only` by visibility db, populating reachability closed WF cache with `A: true` ## Potential risks Worst case scenario would be that reachability returns a false-negative. Due to cache TTL there is the possibility of a false-positive, but eventually the false positive will resolve due to TTL ## Documentation Make sure to add the false-positive caveat to docs. ## Is hotfix candidate? No