deps: upgrade EG Go dependency #1309

mathetake · 2025-10-08T16:32:33Z

Description

This upgrades the Envoy Gateway dependency to the latest main branch commit. Note that the Go level EG dependency is essentially only used in aigw run standalone mode. That is because in the k8s environment, the only code path where we use EG is to generate EG configurations, hence the data structures/control plane/extension server gRPC API which we should assume is stable. Moreover, we still have the e2e test with v1.5, so by no means this implies AIGW main branch cannot work with v1.5. On the other hand, this brings a lot of bug fixes and enhancement to the standalone mode.

mathetake · 2025-10-08T16:44:29Z

go.mod

 // Explicitly specify the Go patch version to be able to purge the CI cache correctly.
 go 1.25.1

-replace go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.61.0 => go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.60.0


fyi we got go run github.com/envoyproxy/ai-gateway/cmd/aigw@main run back

Niiiiiice!!

mathetake · 2025-10-08T17:09:14Z

hmm mcp got broken

mathetake · 2025-10-08T17:25:25Z

<< npx @modelcontextprotocol/[email protected] --cli http://localhost:1975/mcp --transport http --method 'tools/list'
Need to install the following packages:
@modelcontextprotocol/[email protected]
Ok to proceed? (y) y

Failed to connect to MCP server: Error POSTing to endpoint (HTTP 404): unsupported path: /mcp
Failed to connect to MCP server: Error POSTing to endpoint (HTTP 404): unsupported path: /mcp

so looks like the mcp request reaches the ai extproc... i guess some behavioral change around extensionserver

codecov-commenter · 2025-10-08T17:43:40Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77.59%. Comparing base (c6a6937) to head (efefd09).

❌ Your project status has failed because the head coverage (77.59%) is below the target coverage (86.00%). You can increase the head coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1309      +/-   ##
==========================================
+ Coverage   77.55%   77.59%   +0.03%     
==========================================
  Files         123      123              
  Lines       15719    15717       -2     
==========================================
+ Hits        12191    12195       +4     
+ Misses       2900     2896       -4     
+ Partials      628      626       -2

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 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.

mathetake · 2025-10-08T18:40:23Z

/retest

Signed-off-by: Takeshi Yoneda <[email protected]>

mathetake · 2025-10-08T19:48:58Z

cmd/aigw/translate.go

 	}
 	for _, secret := range originalSecrets {
-		mustWriteObj(nil, secret, output)
+		mustWriteObj(&secret.TypeMeta, secret, output)


a bug somehow surfaced with the upgrade

mathetake · 2025-10-08T19:49:22Z

internal/extensionserver/post_translate_modify.go

-			kind, ok := resource.GetStructValue().Fields["kind"]
-			if !ok || kind.GetStringValue() != "HTTPRoute" {
-				continue
-			}


also a bug that surfaced with the upgrade

@zirain this relates to envoyproxy/gateway#7128 (comment), can you help triage this

Yeah won't be an issue for us but sure I think it's a regression on EG side

@mathetake can you raise an issue in EG?

Raised envoyproxy/gateway#7183

@mathetake this might due to kubernetes-sigs/controller-runtime#3229.

it's not from EG side but controller-time.

The fakeclient currently differs from the liveclient in that if a structured object is created that has typemeta set, it will retain that. In contrast to that, the liveclient always strips it. This change makes the fakeclient strip it just like the live client.

nacx

Nice!

go.mod

codefromthecrypt · 2025-10-09T00:52:43Z

EXCELLENT: so this has the fix @nacx made which was very important. func-e inside EG creates an implicit admin server on demand when there is no fixed port one in the yaml. Before, the EG code was broken in how it passed args to cmd Exec (space encoding arg/value in the args slice which is odd worked at all). I can tell by running with --debug on the latest docker image that this bug is now fixed. Not only can you tell the admin server, but it is used to check status of the child process.

aigw         | time=2025-10-09T00:37:55.526Z level=INFO msg="Discovered Envoy admin adminPort" envoyAdminPort=34703
--snip--
aigw         | time=2025-10-09T00:37:59.530Z level=INFO msg="Waiting for Envoy to be ready..." err="unexpected status code: 503"
--snip--
aigw         | [2025-10-09 00:38:11.529][21][debug][http] [source/common/http/conn_manager_impl.cc:2031] [Tags: "ConnectionId":"8","StreamId":"16872216129129905875"] Codec completed encoding stream.
aigw         | time=2025-10-09T00:38:11.529Z level=INFO msg="Envoy is ready!"

**Description** This upgrades the Envoy Gateway dependency to the latest main branch commit. Note that the Go level EG dependency is essentially **only used in aigw run** standalone mode. That is because in the k8s environment, the only code path where we use EG is to generate EG configurations, hence the data structures/control plane/extension server gRPC API which we should assume is stable. Moreover, we still have the e2e test with v1.5, so by no means this implies AIGW main branch cannot work with v1.5. On the other hand, this brings a lot of bug fixes and enhancement to the standalone mode. --------- Signed-off-by: Takeshi Yoneda <[email protected]> Signed-off-by: Hrushikesh Patil <[email protected]>

mathetake force-pushed the depsupgrade2 branch from 4b95a6b to ada9afc Compare October 8, 2025 16:33

mathetake commented Oct 8, 2025

View reviewed changes

mathetake added 4 commits October 8, 2025 12:37

deps: upgrade EG

b9f659e

Signed-off-by: Takeshi Yoneda <[email protected]>

fixes it

93d0d01

Signed-off-by: Takeshi Yoneda <[email protected]>

more

f37cfa8

Signed-off-by: Takeshi Yoneda <[email protected]>

more

efefd09

Signed-off-by: Takeshi Yoneda <[email protected]>

mathetake force-pushed the depsupgrade2 branch from 1db6768 to efefd09 Compare October 8, 2025 19:37

mathetake marked this pull request as ready for review October 8, 2025 19:48

mathetake requested a review from a team as a code owner October 8, 2025 19:48

mathetake commented Oct 8, 2025

View reviewed changes

mathetake changed the title ~~deps: upgrade EG~~ deps: upgrade EG Go dependency Oct 8, 2025

nacx approved these changes Oct 8, 2025

View reviewed changes

aabchoo reviewed Oct 8, 2025

View reviewed changes

go.mod Show resolved Hide resolved

aabchoo approved these changes Oct 8, 2025

View reviewed changes

mathetake merged commit f752c57 into main Oct 8, 2025
47 checks passed

mathetake deleted the depsupgrade2 branch October 8, 2025 21:01

mathetake mentioned this pull request Oct 8, 2025

Standalone mode not working with k8s.io group v1.34.x envoyproxy/gateway#6661

Closed

codefromthecrypt mentioned this pull request Oct 9, 2025

feat: cleanly implement HEALTHCHECK for aigw docker #1314

Merged

deps: upgrade EG Go dependency #1309

deps: upgrade EG Go dependency #1309

Uh oh!

Conversation

mathetake commented Oct 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mathetake Oct 8, 2025

Choose a reason for hiding this comment

Uh oh!

nacx Oct 8, 2025

Choose a reason for hiding this comment

Uh oh!

mathetake commented Oct 8, 2025

Uh oh!

mathetake commented Oct 8, 2025

Uh oh!

codecov-commenter commented Oct 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

mathetake commented Oct 8, 2025

Uh oh!

mathetake Oct 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mathetake Oct 8, 2025

Choose a reason for hiding this comment

Uh oh!

arkodg Oct 9, 2025

Choose a reason for hiding this comment

Uh oh!

mathetake Oct 9, 2025

Choose a reason for hiding this comment

Uh oh!

zirain Oct 9, 2025

Choose a reason for hiding this comment

Uh oh!

arkodg Oct 9, 2025

Choose a reason for hiding this comment

Uh oh!

zirain Oct 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nacx left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

codefromthecrypt commented Oct 9, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

mathetake commented Oct 8, 2025 •

edited

Loading

codecov-commenter commented Oct 8, 2025 •

edited

Loading

mathetake Oct 8, 2025 •

edited

Loading

zirain Oct 11, 2025 •

edited

Loading