K
Q

livenessprobe exec check probe for mongodb is killing pod under k8s 1.2.4

10/27/2016

I'm trying to implement livenessProbes on pods running under K8s. Only very simple probe is working, see example:

    livenessProbe:
      exec:
        command:
        - cat
        - /etc/services
      initialDelaySeconds: 45
      timeoutSeconds: 5

Probe I need to implement is killing pod every 50sec.

    livenessProbe:
      exec:
        command:
          mongo --port 27018 --eval "rs.status()"  | grep  "REMOVED" ; test $? -eq 1
      initialDelaySeconds: 45
      timeoutSeconds: 5

When probe is not implemented and command run inside a pod it is returning 1 if REMOVED string found and 0 if not found as required.

[root@eas-mongo-rs-3-ui81p /]# mongo --port 27018 --eval "rs.status()" | grep "REMOVED" ; test $? -eq 1

Question is, how to implement such command as Container Exec Check.

Regards

-- nikaudio
kubernetes-health-check
mongodb

Similar Questions

Accessing Postgres RDS from Kubernetes cluster in AWS
Sample app for kubernetes 1.2.0
Google Cloud SQL ER_HOST_IS_BLOCKED
In Kubernetes, how to set pods' names when using replication controllers?

1 Answer

2/1/2017

Your command will be quoted and not run as you think, if you want to use pipelines and other more advanced shell functionalities, you can use /bin/sh, a workaround you can also see in the official documentation.

So your probe would look something like this:

 livenessProbe:
  exec:
    command:
      - /bin/sh
      - -c
      - mongo --port 27018 --eval "rs.status()" | grep -vq "REMOVED"
  initialDelaySeconds: 45
  timeoutSeconds: 5

You can also use the -v switch to grep to avoid the reversing of the status with that extra test command.

-- terje
Source: StackOverflow