Status prober should track probes work unit based on PodIP+Port instead of PodIP

[tcnghia]

In what area(s)?

/area networking

/area test-and-release

What version of Knative?

HEAD

Expected Behavior

When multiple Envoys are running on a Pod, a successful probe on one port shouldn't cancel the probing on the others.

Actual Behavior

A successful probe on one port cancel probing on the others.

Steps to Reproduce the Problem

[JRBANCEL]

That's a change in behavior.
Currently, we only validate that the Envoy runs the expected config by probing all known listening ports and considering one successful probe (no matter which port/protocol) a success for that Envoy pod. We do not validate that the config is correct.

You are proposing that probes on all listening ports must succeed. If HTTP and HTTPS are exposed for example and HTTPS is misconfigured, the HTTPS probe will fail forever and IsReady will always be false. Is this what we want?

[tcnghia]

@JRBANCEL In Kourier, I believe each Pod has multiple Envoy containers, in which success in one of them doesn't mean config has propagated to the other. However, we may want to wait for Kourier to come to knative/net-kourier before make this more generic.

[knative-housekeeping-robot]

Issues go stale after 90 days of inactivity.
Mark the issue as fresh by adding the comment /remove-lifecycle stale.
Stale issues rot after an additional 30 days of inactivity and eventually close.
If this issue is safe to close now please do so by adding the comment /close.

Send feedback to Knative Productivity Slack channel or file an issue in knative/test-infra.

/lifecycle stale

[vagababov]

/remove-lifecycle stale

[JRBANCEL]

This will be obsolete once we use Istio VirtualService status instead of probing.

[tcnghia]

@JRBANCEL before VirtualService Status is used, I think we will need this to support your migration to using hidden port on istio-ingressgateway for cluster local traffic?

[tcnghia]

/assign @JRBANCEL

[JRBANCEL]

We don't cancel the probing of a Pod once a probe succeeds since #8795. We wait for all probes to succeed.

Therefore, nothing needs to be done. All the hosts (just one in Istio case) on all the ports will be probed, whether they are on the same Pod or not.

[JRBANCEL]

/close

[knative-prow-robot]

@JRBANCEL: Closing this issue.

In response to this:

/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.