Description
Cadence allows fractional worker_activities_per_second. For instance, worker_activities_per_second=0.167 expected to pass one activity every 6 seconds, but instead some (or most) workflows are not making progress at all.
Steps to Reproduce / How to Trigger
Add the following config to the client' task-list configuration section:
options:
task_list_activities_per_second: 0.167
Then start 10 hello-world workflows (with big enough ScheduleToStart timeouts).
Expected Behavior
Workflows are finished every 6 seconds (they have only one activity): first workflow finished immediately, second one after 6 seconds, etc. All 10 workflows launched at the same time should succeed within 54 seconds.
Actual Behavior
Actual behaviour is flaky. Sometimes multiple workflows succeed, but often it is just two - others are getting timed out. Every time you can notice some internals are skipped: first workflow finished immediately, then second finished in 12 seconds (not in 6), next one in 30 seconds (18s and 24s intervals are like "skipped").
Logs / Screenshots
Environment
- Cadence server version: master:head
- Cadence SDK language and version (if applicable): Language and SDK version
- Cadence web version (if applicable): Web UI version
- DB & version: sqlite, but same happens with Cassandra
- Scale: 1 local instance of cadence-server
Description
Cadence allows fractional worker_activities_per_second. For instance, worker_activities_per_second=0.167 expected to pass one activity every 6 seconds, but instead some (or most) workflows are not making progress at all.
Steps to Reproduce / How to Trigger
Add the following config to the client' task-list configuration section:
Then start 10 hello-world workflows (with big enough ScheduleToStart timeouts).
Expected Behavior
Workflows are finished every 6 seconds (they have only one activity): first workflow finished immediately, second one after 6 seconds, etc. All 10 workflows launched at the same time should succeed within 54 seconds.
Actual Behavior
Actual behaviour is flaky. Sometimes multiple workflows succeed, but often it is just two - others are getting timed out. Every time you can notice some internals are skipped: first workflow finished immediately, then second finished in 12 seconds (not in 6), next one in 30 seconds (18s and 24s intervals are like "skipped").
Logs / Screenshots
Environment