Add --replace flag to intercept command

[petergardfjall]

Description

This PR adds a --replace flag to the telepresence intercept command which, when specified, will replace application containers in the intercepted workload.

This PR builds on the server-side machinery implemented in #3330.

Having the possibility to replace workload containers should provide a solution for tickets like #1646 and #1608.

Checklist

• I made sure to update ./CHANGELOG.yml.
• I made sure to add any docs changes required for my change (including release notes).
• My change is adequately tested.
• I updated DEVELOPING.md with any special dev tricks I had to use to work on this code efficiently.
• I updated TELEMETRY.md if I added, changed, or removed a metric name.
• Once my PR is ready to have integration tests ran, I posted the PR in #telepresence-dev in the datawire-oss slack so that the "ok to test" label can be applied.

[thallgren]

Hi @petergardfjall . Thanks for providing this. Can you please sign your commit and also please fix the lint complaint?

[petergardfjall]

Hi @petergardfjall . Thanks for providing this. Can you please sign your commit and also please fix the lint complaint?

Hi @thallgren, I've addressed your comments.

One thing I noted is that in its present shape --replace won't preserve volume mounts of the original workload pod at the local mount point. I suppose that is to be expected since the replace will simply delete the workload pod (it is implemented by the deleteAppContainer patch if I'm reading it right). I suppose that patch would need to do finer-grained surgery on the application containers if the intercept is to preserve volume mounts. Perhaps something along the lines of:

• replace image with k8s.gcr.io/pause (or similar)
• remove readinessProbe and livenessProbe

That is perhaps a bit more similar in spirit to the swap-deployment command in Telepresence v1?

Either way I think that should not pollute this PR.
If you can provide some guidance I might take a stab at it in a follow-up PR.

[thallgren]

Agree about not polluting this PR. If mounts doesn't work, then that's a separate issue. Not sure why that doesn't work. It really should, even when the container is replaced.

The readinessProbe/livenessProbe intended for the deleted container is also something that needs to be fixed. Not sure about replacing the image with k8s.gcr.io/pause though. The agent really should take over whatever it's doing (including mounts).

[thallgren]

This option is already present in our enterprise edition, so rather than asking you to provide an integration test, I pushed a commit to your branch with the test that we have already have.

[petergardfjall]

Agree about not polluting this PR. If mounts doesn't work, then that's a separate issue.

Opened an issue #3483 with a proposed fix in #3484.