SonarQube Liveness probe failed after helm chart update to version 6.0.1+425

SonarQube
, ,
1

wget ignores no_proxy variable and connects to proxy server. So the liveness probe will fail and the pod will be restarted after a few minutes.

kubectl exec -n sonarqube-enterprise sonarqube-enterprise-app-sonarqube-0 -it -- /bin/bash
> host="$(hostname -i || echo '127.0.0.1')"
> no_proxy=$host
> wget -O- --header="X-Sonar-Passcode: $SONAR_WEB_SYSTEMPASSCODE"  http://${host}:9000/api/system/liveness
> Connecting to my-proxy:3128 (10.0.0.2:3128)
3

Hello @totobaa ,

Welcome to the community!

Could you explain a bit more into details what is the issue you are facing? How do you set http_proxy, https_proxy, no_proxy in the helm chart ?

In your example, what are the value of host and no_proxy and http_proxy ?
Have you tried with export no_proxy=$host as it supposed to be an environment variable?

5

I think the busybox wget ignores the no_proxy variable and send every request to the proxy if configured. It would also send the request the request to the proxy when host → localhost

For the application SonarQube we set these variables in the helm chart

sonarProperties
    http.proxyHost: proxy.example.com
    http.proxyPort: 3128
    https.proxyHost: proxy.example.com
    https.proxyPort: 3128
    http.nonProxyHosts: localhost|127.*|[::1]||<some_internal_domains

We use the PodPreset tool from the RedHat CoP which mounts a config map of proxy variables in each pod.

Environment variables:

HTTPS_PROXY=http://proxy.example.com:3128
HTTP_PROXY=http://proxy.example.com:3128
https_proxy=http://proxy.example.com:3128
http_proxy=http://proxy.example.com:3128
no_proxy=10.10.0.0/21,10.194.48.0/21,localhost,127.0.0.1,169.254.169.254,169.254.170.2,.internal,.eu-central-1.eks.amazonaws.com,kubernetes.default,kubernetes.default.svc,kubernetes.default.svc.cluster,kubernetes.default.svc.cluster.local,.svc,.svc.cluster,.svc.cluster.local
NO_PROXY=10.10.0.0/21,10.194.48.0/21,localhost,127.0.0.1,169.254.169.254,169.254.170.2,.internal,.eu-central-1.eks.amazonaws.com,kubernetes.default,kubernetes.default.svc,kubernetes.default.svc.cluster,kubernetes.default.svc.cluster.local,.svc,.svc.cluster,.svc.cluster.local

I tried it with export no_proxy=$host but it did not work.

Values from my example:
host=10.194.50.171
no_proxy=10.194.50.171
http_proxy=http://proxy.company-name.de:3128

6

Hello @totobaa

Thank you for the additional information.
I don’t believe we fully support the injection of environment variables directly into the pod for the helm chart.
But as you suggested, it seems that busybox wget doesn’t support the no_proxy environment variable.

I have created a ticket to improve our docker image so it would be possible to use this environment variable with wget.

As a workaround, you can:

  • Disable PodPreset for that specific pod, since the traffic within sonarqube will use the values from http.proxyHost and https.proxyHost in sonar.properties
  • Update the liveness probe to disable http proxy for that specific command:
    http_proxy= wget -O- --header="X-Sonar-Passcode: $SONAR_WEB_SYSTEMPASSCODE" http://${host}:9000/api/system/liveness
7

Thank you for the quick reply.

If the pod does not need the proxy variables then I will disable the injection of these variables.