We have 2 KeyCloak clusters running into the same Docker Swarn cluster.
As recommanded we put in place the Prometheus Push Gateway.
How to get both views for JVM indicators :
per instance
per cluster
Actual Behavior
If we do not set PROMETHEUS_GROUPING_KEY_INSTANCE
then we are unable to get a view per cluster because the job's name is the same for the two clusters
If we set PROMETHEUS_GROUPING_KEY_INSTANCE=ENVVALUE:KEYCLOAK_CLUSTER_NAME
then JVM metrics are not correctly taken into account into Prometheus and Grafana, e.g. the JVM memory is really much higher than the truth => back to the issue we tried to solve with using the Prometheus Push Gateway.
Currently we have:
sum(jvm_memory_bytes_used{instance="$instance", area="heap"})*100/sum(jvm_memory_bytes_max{instance="$instance", area="heap"})
Expected Behavior
Be able to define our own custom PROMETHEUS_GROUPING_KEY_JOB in order to separate indicators of different clusters
Description
We have 2 KeyCloak clusters running into the same Docker Swarn cluster. As recommanded we put in place the Prometheus Push Gateway.
How to get both views for JVM indicators :
Actual Behavior
If we do not set PROMETHEUS_GROUPING_KEY_INSTANCE then we are unable to get a view per cluster because the job's name is the same for the two clusters
If we set PROMETHEUS_GROUPING_KEY_INSTANCE=ENVVALUE:KEYCLOAK_CLUSTER_NAME then JVM metrics are not correctly taken into account into Prometheus and Grafana, e.g. the JVM memory is really much higher than the truth => back to the issue we tried to solve with using the Prometheus Push Gateway.
Currently we have: sum(jvm_memory_bytes_used{instance="$instance", area="heap"})*100/sum(jvm_memory_bytes_max{instance="$instance", area="heap"})
Expected Behavior
Be able to define our own custom PROMETHEUS_GROUPING_KEY_JOB in order to separate indicators of different clusters
Environment