Add support for Envoy 1.19.1

[eculver]

#11100 - Adds support in our integration test suite and configuration for testing against Envoy 1.19.x. I'll drop support for 1.15 in a subsequent PR.

All tests are green, but I will go through and try to verify that aspects of what is new in 1.19.x are verified. For example, there was a default added for connect_timeout. Things like this still need to be verified as non-impacting. Also, there are a few new features. I'm wondering if we need to do anything to address/validate them?

[freddygv]

This is looking good, the only thing missing that stands out to me are the website updates (edit: and the changelog entry):
https://www.consul.io/docs/connect/proxies/envoy#supported-versions
https://www.consul.io/commands/connect/envoy#envoy-version

Did you want to do that in the PR that removes 1.15?

[eculver]

@freddygv actually, let me see if removing support for 1.15 works here. I was trying to keep it simple, but given there were no significant changes for 1.19, I can try to remove 1.15 and see how it goes.

[eculver]

@freddygv actually, let me see if removing support for 1.15 works here. I was trying to keep it simple, but given there were no significant changes for 1.19, I can try to remove 1.15 and see how it goes.

Scratch that. I just realized that there are some feature flags that are only relevant for 1.15 so will require some more surgery than I originally thought.

What are your thoughts on having two separate changelog entries? One for adding 1.19 and the other for removing 1.15? And then separate website updates to match?

[freddygv]

That approach sounds good @eculver.

Can you also update the default version listed here?
https://www.consul.io/commands/connect/envoy#envoy-version

[rboyer]

@eculver maybe afterwards you could knock this one out too: #11134

[rboyer]

http: the behavior of the present_match in route header matcher changed. The value of present_match was ignored in the past. The new behavior is present_match is performed when the value is true. An absent match performed when the value is false. Please reference present_match.

We use this for L7 routing and also L7 RBAC

[rboyer]

tcp: switched to the new connection pool by default. Any unexpected behavioral changes can be reverted by setting runtime guard envoy.reloadable_features.new_tcp_connection_pool to false.

👀

[rboyer]

listener: fix a crash which could happen when a filter chain only listener update is followed by listener removal or a full listener update.
yay!

[eculver]

@rboyer did you ever make anything of this:

tcp: switched to the new connection pool by default. Any unexpected behavioral changes can be reverted by setting runtime guard envoy.reloadable_features.new_tcp_connection_pool to false.

Otherwise, I think this is good to merge.

[rboyer]

@eculver

tcp: switched to the new connection pool by default. Any unexpected behavioral changes can be reverted by setting runtime guard envoy.reloadable_features.new_tcp_connection_pool to false.

Nothing for us here, mostly just keeping an eye out for this refactor should something weird happen.

[rboyer]

LGTM

[hc-github-team-consul-core]

🍒 If backport labels were added before merging, cherry-picking will start automatically.

To retroactively trigger a backport after merging, add backport labels and re-run https://circleci.com/gh/hashicorp/consul/463488.

[hc-github-team-consul-core]

🍒 If backport labels were added before merging, cherry-picking will start automatically.

To retroactively trigger a backport after merging, add backport labels and re-run https://circleci.com/gh/hashicorp/consul/465809.

[hc-github-team-consul-core]

🍒❌ Cherry pick of commit c774721 onto stable-website failed! Build Log

[hc-github-team-consul-core]

🍒❌ Cherry pick of commit c774721 onto release/1.10.x failed! Build Log