Hi,
Keycloak starts to having high latencies when there are 50–100 concurrent users
interacting with the Admin REST API. I tested this by using Gatling and run a specific
flow which involves calling the Admin REST API.
Keycloak uses a H2 database during the performance test.
Keycloak is ran in a docker container within docker for Mac.
We are using the following endpoints frequently:
POST /users - to create a new user
GET /roles - to get realm roles
POST /users/{userId}/role-mappings/realm - to assign realm roles to a user
I tried increasing memory in the JVM flags, but that didn’t help.
It shouldn’t be a problem to have that many users right ?
How can I diagnose what’s going on?
Is there a metric endpoint (Prometheus available)?
Are there any other diagnostics I can run?
Greets,
Mark