Pods failing when proxy startup is slow

Hi there,

We have encountered an issue with the envoy-proxy not being live before the main application needs to fetch an external url. In some cases this results in the applications dying. In some cases we have changed our own application to retry for up to a second to give the proxy time to startup, but in other cases we are using 3d party applications. In these cases we can workaround the issue by extending the image with an own image that has a built in wait.

Both solutions are quite ugly and I wonder if there is a more beautiful way to approach this issue.




I know this is an existing issue and has come up before. I don’t believe there is a solution yet, but here are a few links with some discussion.

What exactly you are trying to do when you say applications needs to fetch an external URL ?

I had a similar situation where I have to run “openssl”. I used “until” loop to poll if link is available or not. When it is ready, it runs the command.

Thanks both for the replies. @ericvn I’ll have a look at the other discussions.

@Sourabh_Wadhwa One instance that was failing us, was the jenkins-slave. Jenkins master sets up a slave-pod where the build happens. The slave tries to connect to the master, but fails because the istio-proxy is not running yet. We used a workaround to wrap the slave-image in an other image which would sleep for a certain amount of time before starting the slave. This helps, but it is by far the way we want things to work :slight_smile:. Let’s hope istio comes with a better solution in a next release.