Skip to content

fix: non-blocking server initialization for faster gateway startup #472

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
jchangx wants to merge 2 commits into
base: main
Choose a base branch
from
+39 −14
Open

fix: non-blocking server initialization for faster gateway startup #472

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bcd0e0b
fix: non-blocking server initialization for faster gateway startup
jchangx Apr 7, 2026
a9335c8
fix: keep reloadConfiguration synchronous in dry-run mode
jchangx Apr 9, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 18 additions & 9 deletions pkg/gateway/capabilitites.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,7 @@ import (
"slices"
"strings"
"sync"
"time"

"github.com/google/jsonschema-go/jsonschema"
"github.com/modelcontextprotocol/go-sdk/mcp"
Expand Down Expand Up @@ -110,6 +111,11 @@ func (caps *Capabilities) getResourceTemplateByURITemplate(resource string) (Res
return ResourceTemplateRegistration{}, fmt.Errorf("unable to find resource template")
}

// perServerTimeout is the maximum time to wait for a single server to respond
// during capability listing. This prevents unreachable servers (e.g. VPN-only
// endpoints) from blocking the entire startup for their full transport timeout.
const perServerTimeout = 15 * time.Second

func (g *Gateway) listCapabilities(ctx context.Context, serverNames []string, clientConfig *clientConfig) (*Capabilities, error) {
var (
lock sync.Mutex
Expand All @@ -128,7 +134,10 @@ func (g *Gateway) listCapabilities(ctx context.Context, serverNames []string, cl
// It's an MCP Server
case serverConfig != nil:
errs.Go(func() error {
client, err := g.clientPool.AcquireClient(ctx, serverConfig, clientConfig)
serverCtx, cancel := context.WithTimeout(ctx, perServerTimeout)
defer cancel()

client, err := g.clientPool.AcquireClient(serverCtx, serverConfig, clientConfig)
if err != nil {
log.Logf(" > Can't start %s: %s", serverConfig.Name, err)
return nil
Expand All @@ -137,12 +146,12 @@ func (g *Gateway) listCapabilities(ctx context.Context, serverNames []string, cl

var capabilities Capabilities

tools, err := client.Session().ListTools(ctx, &mcp.ListToolsParams{})
tools, err := client.Session().ListTools(serverCtx, &mcp.ListToolsParams{})
if err != nil {
log.Logf(" > Can't list tools %s: %s", serverConfig.Name, err)
} else {
// Record the number of tools discovered from this server
telemetry.RecordToolList(ctx, serverConfig.Name, len(tools.Tools))
telemetry.RecordToolList(serverCtx, serverConfig.Name, len(tools.Tools))

// Determine the prefix to use for this server's tools
prefix := g.getToolNamePrefix(serverConfig)
Expand All @@ -164,10 +173,10 @@ func (g *Gateway) listCapabilities(ctx context.Context, serverNames []string, cl
}
}

prompts, err := client.Session().ListPrompts(ctx, &mcp.ListPromptsParams{})
prompts, err := client.Session().ListPrompts(serverCtx, &mcp.ListPromptsParams{})
if err == nil {
// Record the number of prompts discovered from this server
telemetry.RecordPromptList(ctx, serverConfig.Name, len(prompts.Prompts))
telemetry.RecordPromptList(serverCtx, serverConfig.Name, len(prompts.Prompts))

for _, prompt := range prompts.Prompts {
capabilities.Prompts = append(capabilities.Prompts, PromptRegistration{
Expand All @@ -178,10 +187,10 @@ func (g *Gateway) listCapabilities(ctx context.Context, serverNames []string, cl
}
}

resources, err := client.Session().ListResources(ctx, &mcp.ListResourcesParams{})
resources, err := client.Session().ListResources(serverCtx, &mcp.ListResourcesParams{})
if err == nil {
// Record the number of resources discovered from this server
telemetry.RecordResourceList(ctx, serverConfig.Name, len(resources.Resources))
telemetry.RecordResourceList(serverCtx, serverConfig.Name, len(resources.Resources))

for _, resource := range resources.Resources {
capabilities.Resources = append(capabilities.Resources, ResourceRegistration{
Expand All @@ -192,10 +201,10 @@ func (g *Gateway) listCapabilities(ctx context.Context, serverNames []string, cl
}
}

resourceTemplates, err := client.Session().ListResourceTemplates(ctx, &mcp.ListResourceTemplatesParams{})
resourceTemplates, err := client.Session().ListResourceTemplates(serverCtx, &mcp.ListResourceTemplatesParams{})
if err == nil {
// Record the number of resource templates discovered from this server
telemetry.RecordResourceTemplateList(ctx, serverConfig.Name, len(resourceTemplates.ResourceTemplates))
telemetry.RecordResourceTemplateList(serverCtx, serverConfig.Name, len(resourceTemplates.ResourceTemplates))

for _, resourceTemplate := range resourceTemplates.ResourceTemplates {
capabilities.ResourceTemplates = append(capabilities.ResourceTemplates, ResourceTemplateRegistration{
Expand Down
26 changes: 21 additions & 5 deletions pkg/gateway/run.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -327,10 +327,6 @@ func (g *Gateway) Run(ctx context.Context) error {
}
}

if err := g.reloadConfiguration(ctx, configuration, nil, nil); err != nil {
return fmt.Errorf("loading configuration: %w", err)
}

// When running in Container mode, disable OAuth notification monitoring and authentication
inContainer := os.Getenv("DOCKER_MCP_IN_CONTAINER") == "1"

Expand Down Expand Up @@ -422,12 +418,32 @@ func (g *Gateway) Run(ctx context.Context) error {
}()
}

log.Log("> Initialized in", time.Since(start))
if g.DryRun {
// In dry-run mode, load capabilities synchronously so we validate
// server configs and report discovered tools before exiting.
if err := g.reloadConfiguration(ctx, configuration, nil, nil); err != nil {
log.Logf("> Initial capability load failed: %s", err)
}
log.Log("> Initialized in", time.Since(start))
log.Log("Dry run mode enabled, not starting the server.")
return nil
}

// Load initial server capabilities in the background. This connects to
// each configured MCP server to discover its tools, prompts, and
// resources. Unreachable servers (e.g. VPN-only endpoints when off-VPN)
// can take 30s+ to time out, so we do this concurrently with starting the
// transport server. The go-sdk's Server.AddTool is thread-safe and
// automatically sends tools/list_changed notifications to connected
// clients once new tools are registered.
go func() {
if err := g.reloadConfiguration(ctx, configuration, nil, nil); err != nil {
Comment on lines +439 to +440

@chatgpt-codex-connector chatgpt-codex-connector Bot Apr 7, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Keep initial reload synchronous in dry-run mode

Launching reloadConfiguration in a goroutine here means gateway run --dry-run can return at line 436 before any capability discovery executes, because dry-run exits immediately and the process can terminate before the background goroutine runs. This regresses dry-run from a real configuration/capability validation pass into a best-effort no-op, so broken/unreachable server configs may now appear successful and expected dry-run output (discovered tools / initialization summary) can be missing.

Useful? React with 👍 / 👎.

log.Logf("> Initial capability load failed: %s", err)
log.Log("> Gateway is running but no tools are available. Tools will load on next configuration update.")
}
log.Log("> Initialized in", time.Since(start))
}()

// Initialize authentication token for SSE and streaming modes
// Skip authentication when running in container (DOCKER_MCP_IN_CONTAINER=1)
transport := strings.ToLower(g.Transport)
Expand Down
Loading