Open kplatis opened 3 days ago
I'm looking into this; thanks for the info @kplatis
Running the service locally isn't too bad, from what I can tell. The RSS stays constant at about ~300Mb, the VIRT gets big though, I assume due to fragmentation.
Running 25 clients in parallel, w/ 2 workers and ~100 requests, I get:
Response time: 6.12 secs
, which isn't great.
However, instrumenting the nexus calls, I see (second column is time in seconds):
count 103.000000
mean 4.922631
std 2.271814
min 0.971000
25% 3.534000
50% 4.746000
75% 6.406500
max 10.432000
So that partly explains the above. I'm wondering, however, if the latency is partly a keycloak auth issue. I will continue to investigate.
I did a performance test locally using the following configuration on Postman:
As it is visible on the following image the CPU of the container maxes out and the memory is not getting released
Also the average response time is ~ 20 seconds which is far from optimal.