Bootstrap flag for connection limit overload manager resource monitor

[seth-epps]

Overview

Fixes #6309

Manual testing

Setting the configuration parameters for the enforced connection limit listener and the connection limits

diff --git a/examples/contour/01-contour-config.yaml b/examples/contour/01-contour-config.yaml
index 6eb7720b..d56b811c 100644
--- a/examples/contour/01-contour-config.yaml
+++ b/examples/contour/01-contour-config.yaml
@@ -184,3 +184,7 @@ data:
     #  socket-options:
     #    tos: 64
     #    traffic-class: 64
+    #
+    omEnforcedHealthListener:
+      address: 0.0.0.0
+      port: 8003
diff --git a/examples/contour/03-envoy.yaml b/examples/contour/03-envoy.yaml
index b9c71f39..6e4e474d 100644
--- a/examples/contour/03-envoy.yaml
+++ b/examples/contour/03-envoy.yaml
@@ -105,6 +105,7 @@ spec:
         - --envoy-cafile=/certs/ca.crt
         - --envoy-cert-file=/certs/tls.crt
         - --envoy-key-file=/certs/tls.key
+        - --overload-dowstream-max-conn=100 # arbitrary value
         command:
         - contour
         image: ghcr.io/projectcontour/contour:main

After running make install-contour-working

# kubectl -n projectcontour port-forward envoy-hq4bj 9001:9001
#
# check listener configs
#
# default stats listner (should ignore -- this is for liveness)
❯ curl -s localhost:9001/config_dump | jq '.configs[] |
  select(."@type" == "type.googleapis.com/envoy.admin.v3.ListenersConfigDump") |
  .dynamic_listeners[] | select(.name == "stats-health").active_state.listener |
  "port: \(.address.socket_address.port_value), ignore_conn_limit: \(.ignore_global_conn_limit)"'
"port: 8002, ignore_conn_limit: true"
#
# enforced connection limit listener (should enforce [null] -- this is for readiness)
#
❯ curl -s localhost:9001/config_dump | jq '.configs[] |
  select(."@type" == "type.googleapis.com/envoy.admin.v3.ListenersConfigDump") |
  .dynamic_listeners[] | select(.name == "health-om-enforced").active_state.listener |
  "port: \(.address.socket_address.port_value), ignore_conn_limit: \(.ignore_global_conn_limit)"'
"port: 8003, ignore_conn_limit: null"
#
# check bootstrap OM config
#
❯ curl -s localhost:9001/config_dump | jq '.configs[] | select(."@type" == "type.googleapis.com/envoy.admin.v3.BootstrapConfigDump") | .bootstrap.overload_manager'
{
  "refresh_interval": "0.250s",
  "resource_monitors": [
    {
      "name": "envoy.resource_monitors.global_downstream_max_connections",
      "typed_config": {
        "@type": "type.googleapis.com/envoy.extensions.resource_monitors.downstream_connections.v3.DownstreamConnectionsConfig",
        "max_active_downstream_connections": "100"
      }
    }
  ]
}

[codecov]

Codecov Report

Attention: Patch coverage is 96.72131% with 4 lines in your changes missing coverage. Please review.

Project coverage is 80.77%. Comparing base (f140c71) to head (45f46d4).
Report is 5 commits behind head on main.

Files with missing lines	Patch %	Lines
cmd/contour/contour.go	0.00%	3 Missing ⚠️
cmd/contour/serve.go	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6794      +/-   ##
==========================================
+ Coverage   80.74%   80.77%   +0.03%     
==========================================
  Files         131      131              
  Lines       19884    19936      +52     
==========================================
+ Hits        16055    16104      +49     
- Misses       3537     3540       +3     
  Partials      292      292              

Files with missing lines	Coverage Δ
cmd/contour/bootstrap.go	100.00% <100.00%> (ø)
cmd/contour/servecontext.go	84.94% <100.00%> (+0.03%)	⬆️
internal/envoy/bootstrap.go	57.14% <100.00%> (+4.87%)	⬆️
internal/envoy/v3/bootstrap.go	93.00% <100.00%> (+0.38%)	⬆️
internal/envoy/v3/stats.go	100.00% <100.00%> (ø)
internal/featuretests/v3/envoy.go	99.20% <100.00%> (+<0.01%)	⬆️
internal/featuretests/v3/featuretests.go	86.90% <100.00%> (+0.04%)	⬆️
internal/provisioner/controller/gateway.go	64.04% <100.00%> (+0.32%)	⬆️
internal/provisioner/model/model.go	100.00% <100.00%> (ø)
...nternal/provisioner/objects/dataplane/dataplane.go	85.78% <100.00%> (+0.03%)	⬆️
... and 4 more

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

The Contour project currently lacks enough contributors to adequately respond to all PRs.

This bot triages PRs according to the following rules:

• After 14d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, the PR is closed

You can:

• Ensure your PR is passing all CI checks. PRs that are fully green are more likely to be reviewed. If you are having trouble with CI checks, reach out to the #contour channel in the Kubernetes Slack workspace.
• Mark this PR as fresh by commenting or pushing a commit
• Close this PR
• Offer to help out with triage

Please send feedback to the #contour channel in the Kubernetes Slack

[github-actions]

The Contour project currently lacks enough contributors to adequately respond to all PRs.

This bot triages PRs according to the following rules:

• After 14d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, the PR is closed

You can:

• Ensure your PR is passing all CI checks. PRs that are fully green are more likely to be reviewed. If you are having trouble with CI checks, reach out to the #contour channel in the Kubernetes Slack workspace.
• Mark this PR as fresh by commenting or pushing a commit
• Close this PR
• Offer to help out with triage

Please send feedback to the #contour channel in the Kubernetes Slack

[github-actions]

The Contour project currently lacks enough contributors to adequately respond to all PRs.

This bot triages PRs according to the following rules:

• After 14d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, the PR is closed

You can:

• Ensure your PR is passing all CI checks. PRs that are fully green are more likely to be reviewed. If you are having trouble with CI checks, reach out to the #contour channel in the Kubernetes Slack workspace.
• Mark this PR as fresh by commenting or pushing a commit
• Close this PR
• Offer to help out with triage

Please send feedback to the #contour channel in the Kubernetes Slack

[sunjayBhatia]

we'll have to add the new omEnforcedHealthListener config file param here:

contour/site/content/docs/main/configuration.md

Line 66 in 74c0c31

## Configuration File

lets also add documentation on the new bootstrap flag here as well:

contour/site/content/docs/main/configuration.md

Line 503 in 74c0c31

### Bootstrap Flags

[sunjayBhatia]

Just messing around with the new health listener, set the overload manager max downstream connections to 1, port-forwarded to the new listener and sent some requests to it in quick succession, could see that the first goes through just fine and a second gets it's connection closed

I also see the failed_updates stat continually increasing, presumably from the readiness checks going to the normal health listener and requests to the admin listener which are both ignoring the global connection limit

[sunjayBhatia]

Just messing around with the new health listener, set the overload manager max downstream connections to 1, port-forwarded to the new listener and sent some requests to it in quick succession, could see that the first goes through just fine and a second gets it's connection closed

I also see the failed_updates stat continually increasing, presumably from the readiness checks going to the normal health listener and requests to the admin listener which are both ignoring the global connection limit

verified the same with the http ingress listener as well

[sunjayBhatia]

Thanks for the changes covering the gateway provisioner!

we dont yet have any API surface in the provisioner to modify readiness/liveness checks on managed contour/envoy instances, but that can be a future feature where we also allow enabling the Overload Manager enforced health endpoint with a readiness check