dapr/dapr · error

flush MCP servers: %w

Error message

flush MCP servers: %w

What it means

Raised while waiting for outstanding MCP server initializations: processor.Flush — the barrier over the shared root loop that resolves when every in-flight init completes — returned an error. Either a pending MCP server (or a component/HTTPEndpoint co-scheduled on the same root loop) failed to init, or the context was canceled before the drain finished.

Source

Thrown at pkg/runtime/mcpservers.go:98

		return nil
	case err := <-ch:
		if err == nil {
			return nil
		}
		err = fmt.Errorf("process MCPServer %s error: %s", s.Name, err)
		if s.Spec.IgnoreErrors {
			log.Errorf("Ignoring error processing MCPServer: %s", err)
			return nil
		}
		log.Warnf("Error processing MCPServer, daprd will exit gracefully: %s", err)
		return err
	}
}

func (a *DaprRuntime) flushOutstandingMCPServers(ctx context.Context) error {
	log.Info("Waiting for all outstanding MCP servers to be processed…")
	if err := a.processor.Flush(ctx); err != nil {
		return fmt.Errorf("flush MCP servers: %w", err)
	}
	log.Info("All outstanding MCP servers processed")
	return nil
}

View on GitHub (pinned to 74ad417027)

Solutions

1. Inspect the init errors logged just before this line to identify the failing MCP server and fix its spec/auth
2. Delay applying MCP servers (or add readiness ordering) until target endpoints and secrets exist
3. Increase terminationGracePeriodSeconds if the failure happens at shutdown drain
4. Remove the offending MCPServer resource and restart daprd

Example fix

# before: MCP server applied before the target endpoint exists
# after: init container / Helm hook waits for the endpoint
startupProbe / preHook: curl -fsS https://mcp.example.com/health

# pre-flight: target MCP endpoints answer before daprd starts
for url in $(kubectl get mcpservers -n <ns> -o json | jq -r '.items[].spec.serverUrl? // empty'); do curl -fsS --max-time 5 "$url" || echo "not ready: $url"; done

if err := rt.flushOutstandingMCPServers(ctx); err != nil && errors.Is(err, context.Canceled) { /* graceful-shutdown drain cut short: extend grace period */ }

When it happens

Trigger: flushOutstandingMCPServers runs during daprd startup/shutdown; it fails when an MCP server init errors (auth failure, unreachable server) or ctx is canceled (shutdown timeout) while inits are still pending.

Common situations: An MCP server pointing at an endpoint that is not yet up blocks the flush; shutdown grace period shorter than slow MCP client construction; concurrent failing components on the shared root loop surfacing through this wrap.

Related errors

• flush http endpoints: %w
• process MCPServer %s error: %s
• server is closed
• flush components: %w
• failed to deserialize MCPServer: %w

AI-assisted analysis of dapr/dapr@74ad417027 (2026-08-16). Data as JSON: /api/errors/050f3f4693acf600. Report an issue: GitHub.