Problem with latency

[Arfey]

Hi ✋

The vast majority of kafka bridge requests have good response times (less than 200ms), but some have more than 2-5 seconds. I tried to find the problem and check the latency from the pod for the /ready handler (to rule out kafka and networking issues), but also some requests had response times more than 1-2 seconds.

Without a load, this problem is not present (produce 2-3 messages per second).

repository: strimzi/kafka-bridge
tag: "0.19.0"
jvmOptions: -Xmx512m -Xms512m

i guess i have not problem with gc or heap (even when i increase the memory it does not give any result). Also i have not problem with cpu:

cpu usage: ~1%
memory: ~400mb

i have discover problem (1-3%) with throttling and increase count of cores for bridge from 1 to 4 (throttling 0%) and it have decreased some count of slow requests but not fully solved my problem

Someone knows what might be wrong or how i can solve a current problem? Thx!

[scholzj]

@sknot-rh @ppatierno Any ideas?

[Torbik]

I'd like to add a few additions to this case:

• the bridge is used as a producer
• thread count (jvm_threads_live_threads) was increasing over time:
  

I suspect so many threads give significant overhead that affects overall performance.

[sknot-rh]

Hi. Are you sure you do use bridge just as a producer? I did some testing and when the bridge is just producing records, it won't get over 54 threads.

[Arfey]

Sure. The current instance is only used to produce messages

[ppatierno]

@Arfey I am going to close this issue, not hearing from you since years. Feel free to re-open if the problem seems to be still in place for you.