Add liveness probe for envoy sidecar container

[rrondeau]

Changes proposed in this PR:

• Add a tcp liveness probe to envoy container

How I've tested this PR: I got this patch on my consul-k8s fork running for months

How I expect reviewers to test this PR: ¯_(ツ)_/¯

Checklist:

• [x] Tests added
• [ ] CHANGELOG entry added (HashiCorp engineers only, community PRs should not add a changelog entry)

[lkysow]

@rrondeau I wonder if this is necessary now with the new health checks support?

[rrondeau]

@lkysow hum i dont think the new health checks support and my change have the same goal. This change add a k8s liveness probe to the consul-connect-envoy-sidecar, not a check for consul connect. This is to tell k8s to restart the envoy container if the public listener is not listening on its port for whatever reason.

[lkysow]

Ahh sorry, you're right. Have you seen the envoy container not bind to that port? I'm wondering if there's an underlying issue.

[rrondeau]

Have seen it but it was due to temporary network issue between envoy and its consul client. Envoy was not receiving any clusters/listeners. I think having liveness probe on a container is a best practice and can always be useful.

This probe also helped me when something broke my consul config injection like i said in #383

[rrondeau]

@lkysow @ishustava any updates ?

[hashicorp-cla]

All committers have signed the CLA.

[rrondeau]

Just saw that this had been added in the 1.0 version by @ishustava ! Thanks !