Openshift service is not available after short inactivity

Question

We have our project hosted in OpenShift (OKD to be precise. We host it ourselves). The setup is as follows:

• Routing server (Spring Boot 1.5.8 with Zuul): This one takes all the incoming traffic and routes it to the correct services

• Multiple services (all with Spring Boot): Here is all the business logic

We use SOAP for calling other services in this project.

Currently, when we call the application, the call goes to the routing server, which then routes it to the main business service. After a short inactivity of about one hour, our main business service is not reachable via the external call. The edge server however is available and callable 100% of the time. We do get a 504 Gateway Timeout exception from the system when we call it. We already figured out that this is the timeout of the route in openshift (haproxy.router.openshift.io/timeout in the route).

The core problem is, that OpenShift seems to hibernate the main business service after an inactivity of about one hour. After a delay of 15 minutes however the calls seem to find their destination and the data gets processed correctly.

How can we turn this behaviour off?

Edit 1:

• We have the same application in normal "old fashioned" VMs in production. We don't have any problems there.
• We noticed that the services can be "kept alive" when we call them regulary. We built a small service which calls theme regulary (every 15 min). This way it seems to work. But this is not a production ready workaround IMO.

Edit 2:

Our pod config (some names are anonymized):

https://gist.github.com/moritzluedtke/6867499b0acbb2d7b5a9a70e49b0d45c

We do not use autoscaler.

Edit 3:

Our deployment configs (some names are anonymized):

https://gist.github.com/moritzluedtke/dc7c1078fe9cc7e4aeb737094849fc1b

OpenShift Master:      v3.11.0+1c3e643-87
Kubernetes Master:     v1.11.0+d4cacc0
OpenShift Web Console: v3.11.0+ea42280

Edit 4:

It seems that this is not a problem with OpenShift but rather our tech stack. I will update this question, as soon as we have a solution.

Vander

Please provide your pod and your autoscaler config – Egalitarian

I added the config. Thanks for looking into this. – Vander

OpenShift does not hibernate or sleep services. What you're seeing is more likely an issue with the Routing server. I would monitor the logs of the Routing server pods for anything suspicious. I notice you deployed a pod directly, you don't want to do that. You should deploy your application workloads in a DeploymentConfig object fronted by a Service. – Decontaminate

Hey Nick, thanks for the input! We will try that asap. I just want to add that we do have a deployment config in place. Ariel just asked for the pod config so I provided that. I will add the deployment config in a minute. – Vander

Answer 1

We ended up updating the database driver (POSTGRESQL) and switching out the default connection pool from spring.

pom.xml:

<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-data-jpa</artifactId>
    <exclusions>
        <exclusion>
            <groupId>org.apache.tomcat</groupId>
            <artifactId>tomcat-jdbc</artifactId>
        </exclusion>
    </exclusions>
</dependency>
<dependency>
    <groupId>com.zaxxer</groupId>
    <artifactId>HikariCP</artifactId>
    <version>${HikariCP.version}</version>
</dependency>

Vander