mirror of
https://github.com/hanzoai/tasks.git
synced 2026-10-30 14:56:19 +00:00
Vladyslav SimonenkoandGitHub
681e5107c7
## What changed? Changes activity metrics ### Deprecated Metrics - **activity_e2e_latency** → Deprecated, as it measures individual attempts rather than true end-to-end latency, and replaced with clearer **activity_start_to_close_latency** ### New & Replacement Metrics - **activity_start_to_close_latency**: Measures the latency from activity start to close (per attempt). - **activity_schedule_to_close_latency**: Measures true end-to-end duration, including retries and backoff. - **activity_success**: Counts the number of succeeded activities. Aligned with the existing **workflow_success** counter. - **activity_fail**: Counts final failures for activities. Similar to **workflow_failed,** although it doesn’t include retries. - **activity_timeout**: Incremented on the final activity timeout (including ScheduleToStartTimeout), tagged by `timeout_type`. Aligned with the existing **workflow_timeout** counter. - **activity_task_fail**: Counts failures for activities including retries. Note that we don’t need to capture the number of retries, as this metric represents this number well. - **activity_task_timeout**: Incremented on the activity attempt timeout (including ScheduleToStartTimeout), tagged by `timeout_type`. - **activity_cancel:** Incremented when an activity is cancelled **** ## Why? ActivityE2ELatency is inaccurate, since it measures only individual attempts. The other metrics are not recorded yet. ## How did you test it? - [X] built - [X] run locally and tested manually - [X] covered by existing tests - [X] added new unit test(s) - [ ] added new functional test(s) Ran the server locally and checked the metrics in Prometheus: Test 1: Activity with 5 attempts (1s each, 1s gap, all time out) - activity_task_timeout = 5 - activity_timeout = 1 - activity_success = 0 - activity_fail = 0 (counted as timeout instead) - activity_task_fail = 0 (counted as timeout instead) - activity_schedule_to_close_latency_sum = 9s - activity_end_to_end_latency_sum = 1s (deprecated) - activity_start_to_close_latency_sum = 1s Test 2: Basic helloworld activity - activity_success = 1 Test 3: Activity with 5 attempts (1s each, 1s gap, all fail except for the last one) - activity_task_fail = 4 - activity_success = 1