High CPU observed with Multicast Aggregate pattern

[Lakshmi Panala <lp...@outlook.com>]

Hello All,

I understand there was a fix for high CPU utilization issue with Camel Multicast Aggregation pattern. The fix was to increase the delay time from 1 nanosecond to 1001 nanoseconds.  https://issues.apache.org/jira/browse/CAMEL-11750

I’m seeing a similar issue. We are using Camel v2.25.x for an application that makes parallel calls using multicast DSL. One of the parallel calls takes significantly loner time than the other. When this happens, MulticastProcessor-AggregateTask thread seems to be consuming more CPU. The CPU consumption itself is about 3.5% per MulticastProcessor-AggregateTask thread. There are 10 such threads each consuming about same amount of CPU pushing the overall CPU usage to more than 30%. The response time from the backend for the long running task is about 3 sec at 95th percentile. I’ve 30 threads in the multicast thread pool. If I reduce the timeout for the backend call from 3sec to 1 sec, CPU usage does come down, but that means more failures to the consumers.


Wondering if you have any suggestions to address this issue. Is there any config param that I can pass to Camel multicast DSL to set the delay so that it doesn’t eat the CPU every 1001 nanoseconds or are there any other solutions? Please advise. 


Thank you.
Lakshmi