Hi @Maxim_Chuvilyaev
you might be running into a bottleneck with the number of task queue partitions.
Task queue partitions can be set via “matching.numTaskqueueWritePartitions” and “matching.numTaskqueueReadPartitions” in dynamic config.
Dynamic config for docker compose
Dynamic config for helm charts
Dynamic config knobs on server side:
read
write
Without knowing the number of workflow tasks you are running / second it’s hard to tell what number you could tune that to, but try updating the default 4 to 10 for example and see if that makes a difference for you.
On the server metrics side I think you can also look at all the different matching service metrics. poll_success_sync and poll_success should give you and idea of the overall sync-match rate and then compare it when you tune the numbers.
Hope this helps.