[keycloak-user] Fine tuning KC for performance

[Upananda Singha]

Hi All,

Could anybody please point out whats the Keycloak load bearing capacity in
terms of TPS (transactions per sec).
As I already indicated I have been using 2 instances of Keycloak connected
to shared PostgreSQL Database. (Keycloak instances being
clustered using standalone-ha configuration).

I have been trying Reset Pwd api call on a 500 user base.
I am running Jmeter test suit to reset pwd and trying to post in a rate
around 100 msg per sec.
But as per Jemeter thoroughput results I can see max around thoroughput 80
which seems really less when I am having 2 instances of KCs.
My load balancer is doing a round robin scheduling for forwarding the
requests to the KC instances.
Can anybody share his/her experience regarding the load test results for
Keycloak? Any information will be highly appreciated.

Any KC optimization configuration would be great...

Thanks & Regds
Upananda Singha


On Wed, Feb 21, 2018 at 12:22 PM, Upananda Singha <upananda313 at gmail.com>
wrote:

> Hi All,
>
> I am trying to test a KC cluster (with 2 KC nodes) using standalone-ha
> mode.
> I have configured a shared database of PostgreSQL (2ndQuadrant with BDR -
> having 2 Database nodes and pointing KC nodes pointing to single node).
>
> Each of the KC nodes is configured with
> JAVA_OPTS="-*Xms256m* -*Xmx1024m* -XX:MetaspaceSize=96M
> -XX:MaxMetaspaceSize=256m -Djava.net.preferIPv4Stack=true"
>
> And default connection pool size set to 40 each.
>
> But when I run traffic I don't see combined throughput going beyond 80/90
> per sec
> when traffic hits both the KC nodes in a round robin fashion.
>
> Can anybody give some idea what kind of tuning I can try to increase the
> throughput. With 2 nodes of KC we are looking for  throughput at least
> 200-300.
>
> Thanks & Regds,
> Upananda
>
>