Add startup timeout to prevent gateway hanging on unresponsive MCP servers by Copilot · Pull Request #477 · github/gh-aw-mcpg

Copilot · 2026-01-26T14:49:23Z

Gateway would hang indefinitely during initialization if any backend MCP server failed to respond, blocking all other backends from starting.

Changes

Configuration: Added startupTimeout field (default: 60s) to gateway config with package-level constants (DefaultStartupTimeout, DefaultToolTimeout, DefaultPort).

Timeout enforcement: Implemented goroutine + buffered channel pattern in GetOrLaunch() and GetOrLaunchForSession():

// Launch connection in goroutine with buffered channel to prevent leaks
resultChan := make(chan connectionResult, 1)
go func() {
    conn, err := mcp.NewConnection(l.ctx, cmd, args, env)
    resultChan <- connectionResult{conn, err}
}()

// Timeout via select instead of context.WithTimeout to preserve connection after startup
select {
case result := <-resultChan:
    return result.conn, result.err
case <-time.After(l.startupTimeout):
    return nil, fmt.Errorf("server startup timeout after %v", l.startupTimeout)
}

Graceful degradation: Failed backends log warnings and gateway continues initialization with remaining backends.

Technical decisions

Goroutine + time.After() instead of context.WithTimeout to avoid canceling connection after successful startup
Buffered channel (size 1) allows goroutine to complete without blocking if timeout occurs first
Enhanced error messages with actionable suggestions to increase timeout if needed

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

nonexistent.local
- Triggering command: /tmp/go-build3377049250/b269/launcher.test /tmp/go-build3377049250/b269/launcher.test -test.testlogfile=/tmp/go-build3377049250/b269/testlog.txt -test.paniconexit0 -test.timeout=10m0s -test.v=true go user.email t[bot] (dns block)
- Triggering command: /tmp/go-build481144272/b273/launcher.test /tmp/go-build481144272/b273/launcher.test -test.testlogfile=/tmp/go-build481144272/b273/testlog.txt -test.paniconexit0 -test.timeout=10m0s -test.v=true dleware.test /tmp/go-build3677275879/b097/ ortcfg.link . --gdwarf2 --64 GL0vaMStIAuGbawbr5/sWo5QoaFO32xMT3YHtp6/k1XMsJiotFONAaFTgZJc (dns block)
- Triggering command: /tmp/go-build3502506537/b273/launcher.test /tmp/go-build3502506537/b273/launcher.test -test.testlogfile=/tmp/go-build3502506537/b273/testlog.txt -test.paniconexit0 -test.timeout=10m0s -test.v=true -lang=go1.25 docker-buildx x_amd64/vet se 7049250/b028/vetls-files x_amd64/cgo x_amd64/vet 791 6791 791 aw-mcpg/awmg se 7049250/b141/vet-c x_amd64/vet aw-mcpg/awmg (dns block)
this-host-does-not-exist-12345.com
- Triggering command: /tmp/go-build3377049250/b278/mcp.test /tmp/go-build3377049250/b278/mcp.test -test.testlogfile=/tmp/go-build3377049250/b278/testlog.txt -test.paniconexit0 -test.timeout=10m0s -test.v=true g_.a --global rgo/bin/bash credential.helpe/opt/hostedtoolcache/go/1.25.6/x64/pkg/tool/linux_amd64/vet (dns block)

If you need me to access, download, or install something from one of these locations, you can either:

Configure Actions setup steps to set up my environment, which run before the firewall is enabled
Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)

Original prompt

This section details on the original issue you should resolve

<issue_title>gateway need to handle dangling mcp servers</issue_title>
<issue_description>See errors: github/gh-aw#11814</issue_description>

Comments on the Issue (you are @copilot in this section)
@pelikhan https://github.com/github/gh-aw/pull/11817

Fixes gateway need to handle dangling mcp servers #469

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

Use goroutine with timeout channel instead of context.WithTimeout to prevent connection from being canceled after successful startup. Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

Eliminates code duplication by defining connectionResult once at the package level instead of duplicating it in both GetOrLaunch() and GetOrLaunchForSession().

The buffered channel with size 1 allows the goroutine to complete even if timeout occurs first, preventing goroutine leaks. Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

Replace magic numbers with package-level constants (DefaultStartupTimeout, DefaultToolTimeout, DefaultPort) to maintain consistency and avoid duplication. Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

Initial plan

9778bc3

Copilot AI assigned Copilot and lpcox Jan 26, 2026

Copilot started work on behalf of lpcox January 26, 2026 14:49 View session

Copilot AI and others added 7 commits January 26, 2026 14:53

Add startup timeout for MCP backend servers to handle dangling servers

107cc4f

Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

Add tests for startup timeout configuration

a1140ca

Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

Fix linting error in common_test.go (pre-existing issue)

5282bb0

Fix timeout implementation to not cancel connection after startup

a334921

Use goroutine with timeout channel instead of context.WithTimeout to prevent connection from being canceled after successful startup. Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

Address code review: extract connectionResult type to package level

2d91b62

Eliminates code duplication by defining connectionResult once at the package level instead of duplicating it in both GetOrLaunch() and GetOrLaunchForSession().

Add comments clarifying buffered channel prevents goroutine leaks

f84a8d5

The buffered channel with size 1 allows the goroutine to complete even if timeout occurs first, preventing goroutine leaks. Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

Use config constants for default timeout values

7f1fc67

Replace magic numbers with package-level constants (DefaultStartupTimeout, DefaultToolTimeout, DefaultPort) to maintain consistency and avoid duplication. Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

Copilot AI changed the title ~~[WIP] Fix gateway to handle dangling mcp servers~~ Add startup timeout to prevent gateway hanging on unresponsive MCP servers Jan 26, 2026

Copilot AI requested a review from lpcox January 26, 2026 15:13

Copilot finished work on behalf of lpcox January 26, 2026 15:13

lpcox marked this pull request as ready for review January 26, 2026 15:39

lpcox merged commit 25f602d into main Jan 26, 2026
3 checks passed

lpcox deleted the copilot/fix-dangling-mcp-servers branch January 26, 2026 15:39

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add startup timeout to prevent gateway hanging on unresponsive MCP servers#477

Add startup timeout to prevent gateway hanging on unresponsive MCP servers#477
lpcox merged 8 commits intomainfrom
copilot/fix-dangling-mcp-servers

Copilot AI commented Jan 26, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

Copilot AI commented Jan 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Technical decisions

I tried to connect to the following addresses, but was blocked by firewall rules:

Comments on the Issue (you are @copilot in this section)

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Copilot AI commented Jan 26, 2026 •

edited

Loading