I have two applications - app1 and app2, where app1 is a config server that holds configs for app2. I have defined /readiness endpoint in app1 and need to wait till it returns OK status to start up pods of app2.
It's crucial that deployment of app2 wait till kubernetes receives Http Status OK from /readiness endpoint in app1 as it's a configuration server and holds crucial configs for app2.
Is it possible to do this kind of deployment dependency?
You can use initContainers. Following is an example of how you can do in your YAML file
initContainers:
- name: wait-for-other-pod
image: docker.some.image
args:
- /bin/sh
- -c
- >
set -x;
while [ $(curl -sw '%{http_code}' "http://www.<your_pod_health_check_end_point>.com" -o /dev/null) -ne 200 ]; do
sleep 15;
done
I have used curl to hit the health check endpoint, you can use any other UNIX command to check if the other pod is ready.
If you have a dependency on k8s resources, you can make use of stackanetes/kubernetes-entrypoint example:
initContainers:
- command:
- kubernetes-entrypoint
name: init-dependency-check
env:
- name: POD_NAME
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.name
- name: NAMESPACE
valueFrom:
fieldRef:
apiVersion: v1
fieldPath: metadata.namespace
- name: DEPENDENCY_SERVICE
- name: DEPENDENCY_DAEMONSET
- name: DEPENDENCY_CONTAINER
- name: DEPENDENCY_POD_JSON
value: '[{"labels":{"app.kubernetes.io/name":"postgres"}}]'
- name: COMMAND
value: echo done
image: projects.registry.vmware.com/tcx/snapshot/stackanetes/kubernetes-entrypoint:latest
securityContext:
privileged: true
runAsUser: 0
In the above example, the pod with initContainer init-dependency-check will wait until pod with label "app.kubernetes.io/name":"postgres" is in the Running state. Likewise you can make use of DEPENDENCY_SERVICE, DEPENDENCY_DAEMONSET, DEPENDENCY_CONTAINER
Yes, it's possible using Init Containers (also, see this blog post for some background re timing) but a better, more Kubernetes-native pattern is to use retries and timeouts rather than hard-coding dependencies in this way.
initContainers:
- name: wait-for-dependent-service
image: stefanevinance/wait-for-200
env:
- name: URL
value: http://dependent-service.{{.Release.Namespace}}.svc.cluster.local:3000
In my case the application don't have a /readiness endpoint. We have master pods and worker pods and we want the worker pods to start up only after master pod is up and running. The master pod runs an application listening on TCP port 80. I ended up using netcat in an init container in the worker pod.
nc -z <host> <port>
nc will check if the port is open, exiting with 0 on success, 1 on failure.
initContainers:
- name: wait-for-master-before-starup
image: busybox
command: ["sh", "-c", "until nc -z master-service 80 > /dev/null; do echo Waiting for master.; sleep 2; done;"]
If you already have a health/readiness endpoint, its much better to use it like mentioned in other answers here.
Using wait-for-it.sh is actually quite easy:
initContainers:
- name: wait-for-app1
image: image-docker-containing-sh
args:
- /bin/sh
- -c
- /usr/app/wait-for-it.sh app1:<portapp1> -t 0
Of course retries and timeouts are the way to go, but this works great as a workaround.
same as @Vishrant Answer... just other way to write bash commands in single line in YAML file
initContainers:
- name: wait-for-some-pod
image: yourDockerImage
command: ["/bin/sh","-c"]
args: ['while [ $(curl -ksw "%{http_code}" "https://<pod_health_check_end_point>" -o /dev/null) -ne 200 ]; do sleep 5; echo "health check failed . Waiting for the service..."; done']
