ProcessDefinitionResourceImpl.submitForm taking more time

Camunda 7 Topics Discussion & Questions
1

Hi Team, I am using Caumda 7.11 version and having 6 Camunda pods running in production. when creating a new process instance with rest api taking long time. We are using Signoz to trace the details. please find the screen shot below

Screenshot from 2023-03-22 13-32-52
Screenshot from 2023-03-22 13-32-52873×492 27.7 KB

I could see clearly first query(ACT_GE_BYTEARRAY TABLE) took 0.41ms and next query(ACT_RE_PROCDEF TABLE) took 0.6ms , but 0.41ms to 0.6ms there is a gap of 500ms approximately. So kindly let us know why this much time taken to trigger the next query and also please provide us the resolution

Thanks in advance.

2

Hi All, Is there any update on query? It would be great help to me if you guys can provide me with resolution or some inputs.

Thanks in advance.

3

Hi @aniinprni,

in which context do you trace the database? What is happening at the end user?

The reason for this delay could be

  • lack of computing power in the process engine
  • Garbage collection in the JVM
  • overloaded network and dropped TCP packages
  • slow browser running expensive Javascript programs in other Tabs
  • maybe more

Hope this helps, Ingo

4

Hi @Ingo_Richtsmeier, Thank you for your response. Below are the some more details on Camunda pods

  • Each pod having 2GB,2 cpus
  • Always pod memory is having 700-800 MB
  • We are using only REST API’s, It has nothing to do with UI
  • Very less traffic as well.

It would be a great help, if you could enlighten us with more details on Overloaded network and dropped TCP packages and maybe more.

Thanks

5

Hi @aniinprni,

In that case, I would start to inspect the default behavior in a (local) test system with the root log level set to debug.

You will get a lot of information and take a close look if there is anything special happening in between the two select statements.

I recently read a story in a German magazine about Observability in DevOps and the author shared a story about users complaining about a slow database. But the metrics on the database were fine, and the packages could not be transferred over the saturated network, which led to poor response times.

I don’t know how to monitor the tcp traffic, maybe you can find another expert on this topic.

Hope this helps, Ingo

6

Hi Thank you for your response. We will look into this and get back to you.

7