Increase limit for number of parallel fetchAndLock calls?

Hi, we are currently using Camunda 7.16 (community edition) and are running into an issue with the hardcoded cap of 200 parallel pending fetchAndLock calls per process engine instance.
Is it possible/sensible to raise that cap to a higher number (as an example 2000)? If so, what would be needed to trigger such a feature request?

Background:

  1. We are forced by other business requirements to limit the number of requested topicnames per fetchAndLock. So we cannot reduce the number of fetchAndLock calls by requesting more topicnames per call.
  2. Due to FetchAndLock long polling response times are inconsistent with multiple instances of Camunda Engine (raised by one of my colleagues back then) we are forced to send all fetchAndLock calls to one process engine instance only.

Obviously other suggestions to address our problem are also welcome.

1 Like

Hi @camunda,

i was wondering if there is the possibility to make the hardcoded cap for the queue configurable?
Seen this question already several times in the forum, and since the external task pattern is getting popular, maybe we should think about it?

Cheers,
Patrick

Hello @patrick.schalk and @Elnubnub ,

the respective line of code can be found here:

Fun Fact: This impl is provided by an SPI, so you can override it for yourself. By extending this class, you will just have to set the field queue to a queue with your desired size.

The SPI can be found here:

If you want this to be changed, please feel free to file a an issue and a pull request (or maybe search for one that already exists for this issue and contribute).

I hope this helps

Jonathan

1 Like