[OTAGENT-1149] Fix standalone otel-agent remote-config/configsync guards leaking from env vars (#54485)

## Summary
- Standalone `otel-agent` (`DD_OTEL_STANDALONE=true`) disables `remote_configuration.enabled` at `SourceFile` priority, which is lower than `SourceEnvVar`. A deployment tool that colocates `otel-agent` with a core Datadog Agent — e.g. the Datadog Operator's DaemonSet path — injects `DD_REMOTE_CONFIGURATION_ENABLED=true` into every container including `otel-agent`, silently overriding the guard. This crashes the embedded trace agent's RC client at startup (exit 255: `could not instantiate the tracer remote config client: grpc client disabled via cmd_port: -1`), since standalone mode has already disabled the core-agent IPC (`cmd_port=-1`) the RC client would need.
- Raise the guard to `SourceAgentRuntime`, which outranks `SourceEnvVar`, so it can no longer be overridden by an injected env var.
- Apply the same treatment to configsync: force `agent_ipc.config_refresh_interval=0` under standalone mode, so a colocated core agent's `DD_AGENT_IPC_CONFIG_REFRESH_INTERVAL` env var can't make "standalone" `otel-agent` keep syncing config from a core agent it isn't supposed to depend on.

See [OTAGENT-1149](https://datadoghq.atlassian.net/browse/OTAGENT-1149) for the full investigation (verified end-to-end against the Datadog Operator in a kind cluster).

## Test plan
- [x] Added `TestStandaloneModeIgnoresCoreAgentIPCEnvVars` reproducing the env-var-leak scenario (`cmd/otel-agent/config/agent_config_test.go`)
- [x] `dda inv test --targets=./cmd/otel-agent/config/...` passes
- [x] `dda inv otel-agent.build` succeeds
- [x] Re-verified end-to-end against the Datadog Operator DaemonSet manifest from the ticket: built a local Linux/arm64 image with the fix, deployed via the Datadog Operator into a kind cluster with `remoteConfiguration.enabled: true` (the previously-crashing config), and confirmed `otel-agent` starts cleanly and stays up (0 restarts). Logs show `configsync disabled: agent_ipc.config_refresh_interval invalid: 0`, no RC-client crash, and the expected standalone boot sequence (`Starting dogtelextension in standalone mode`, tagger gRPC server on :15000), with OTLP receivers up and payloads posting successfully to the Datadog API.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

[OTAGENT-1149]: https://datadoghq.atlassian.net/browse/OTAGENT-1149?atlOrigin=eyJpIjoiNWRkNTljNzYxNjVmNDY3MDlhMDU5Y2ZhYzA5YTRkZjUiLCJwIjoiZ2l0aHViLWNvbS1KU1cifQ

Co-authored-by: yang.song <yang.song@datadoghq.com>

Author: songy23

cmd/otel-agent/config/agent_config.go

@@ -179,6 +179,34 @@ func NewConfigComponent(ctx context.Context, ddCfg string, uris []string) (confi
 	fmt.Printf("setting log level to: %v\n", logLevelReverseMap[activeLogLevel])
 	pkgconfig.Set("log_level", logLevelReverseMap[activeLogLevel], pkgconfigmodel.SourceFile)
 
+	// Standalone mode runs without a core Datadog Agent on the same host, so
+	// every client that would otherwise contact it over IPC (trace-agent
+	// hostname acquisition, remote tagger, remote workloadmeta, configsync,
+	// ...) must be disabled. cmd_port=-1 is the conventional way to express
+	// "no core agent IPC" and is honored by those callers; forcing it here
+	// means users only have to set DD_OTEL_STANDALONE=true.
+	//
+	// All of these are set with SourceAgentRuntime, which outranks
+	// SourceEnvVar, so they can't be silently re-enabled by a deployment tool
+	// that colocates otel-agent with a core agent and injects that agent's
+	// env vars (e.g. DD_REMOTE_CONFIGURATION_ENABLED=true,
+	// DD_AGENT_IPC_CONFIG_REFRESH_INTERVAL) into this container too.
+	//
+	// This must run before the getDDExporterConfig call below, since a
+	// standalone config without a Datadog exporter (a supported shape) makes
+	// that call return early via ErrNoDDExporter, which would otherwise skip
+	// these guards entirely.
+	if pkgconfig.GetBool("otel_standalone") {
+		pkgconfig.Set("cmd_port", -1, pkgconfigmodel.SourceAgentRuntime)
+		// There is no core agent to sync config from; disable configsync
+		// regardless of whether it's configured over agent_ipc.port or
+		// agent_ipc.use_socket.
+		pkgconfig.Set("agent_ipc.config_refresh_interval", 0, pkgconfigmodel.SourceAgentRuntime)
+	}
+	if pkgconfig.GetInt("cmd_port") <= 0 {
+		pkgconfig.Set("remote_configuration.enabled", false, pkgconfigmodel.SourceAgentRuntime)
+	}
+
 	ddc, err := getDDExporterConfig(cfg)
 	if err == ErrNoDDExporter {
 		return pkgconfig, err
@@ -224,18 +252,6 @@ func NewConfigComponent(ctx context.Context, ddCfg string, uris []string) (confi
 	if addr := ddc.Traces.Endpoint; addr != "" {
 		pkgconfig.Set("apm_config.apm_dd_url", addr, pkgconfigmodel.SourceFile)
 	}
-	// Standalone mode runs without a core Datadog Agent on the same host, so
-	// every client that would otherwise contact it over IPC (trace-agent
-	// hostname acquisition, remote tagger, remote workloadmeta, ...) must be
-	// disabled. cmd_port=-1 is the conventional way to express "no core agent
-	// IPC" and is honored by those callers; forcing it here means users only
-	// have to set DD_OTEL_STANDALONE=true.
-	if pkgconfig.GetBool("otel_standalone") {
-		pkgconfig.Set("cmd_port", -1, pkgconfigmodel.SourceAgentRuntime)
-	}
-	if pkgconfig.GetInt("cmd_port") <= 0 {
-		pkgconfig.Set("remote_configuration.enabled", false, pkgconfigmodel.SourceFile)
-	}
 
 	if !pkgconfig.IsConfigured("apm_config.features") {
 		apmConfigFeatures := []string{}

cmd/otel-agent/config/agent_config_test.go

@@ -207,6 +207,27 @@ func (suite *ConfigTestSuite) TestAgentConfigWithDatadogYamlKeysAvailable() {
 	assert.Equal(t, 60, c.GetInt("agent_ipc.config_refresh_interval"))
 }
 
+// TestStandaloneModeIgnoresCoreAgentIPCEnvVars reproduces OTAGENT-1149: a
+// deployment tool that colocates otel-agent with a core Datadog Agent (e.g.
+// the Datadog Operator's DaemonSet) injects that agent's env vars —
+// DD_REMOTE_CONFIGURATION_ENABLED=true and DD_AGENT_IPC_CONFIG_REFRESH_INTERVAL
+// — into the otel-agent container too. Standalone mode must win regardless,
+// since there is no core agent IPC endpoint for the RC client or configsync
+// to talk to.
+func (suite *ConfigTestSuite) TestStandaloneModeIgnoresCoreAgentIPCEnvVars() {
+	t := suite.T()
+	t.Setenv("DD_OTEL_STANDALONE", "true")
+	t.Setenv("DD_REMOTE_CONFIGURATION_ENABLED", "true")
+	t.Setenv("DD_AGENT_IPC_CONFIG_REFRESH_INTERVAL", "60")
+
+	c, err := NewConfigComponent(context.Background(), "", []string{"testdata/config.yaml"})
+	require.NoError(t, err)
+
+	assert.Equal(t, -1, c.GetInt("cmd_port"))
+	assert.False(t, c.GetBool("remote_configuration.enabled"))
+	assert.Equal(t, 0, c.GetInt("agent_ipc.config_refresh_interval"))
+}
+
 func (suite *ConfigTestSuite) TestAgentConfigSetAPMFeaturesFromDatadogYaml() {
 	t := suite.T()
 	fileName := "testdata/config_default.yaml"

cmd/otel-agent/subcommands/run/command.go

@@ -313,7 +313,9 @@ func standaloneAgentFxOptions(params *cliParams) fx.Option {
 		// Resolve hostname locally; no core agent to ask
 		hostnameimpl.Module(),
 		// No on-init config sync (no core agent to sync from); periodic sync is also
-		// effectively disabled by the default agent_ipc.config_refresh_interval=0
+		// force-disabled in agent_config.go (agent_ipc.config_refresh_interval=0 via
+		// SourceAgentRuntime) so it can't be re-enabled by an env var meant for a
+		// colocated core agent.
 		configsyncfx.Module(configsync.NewParams(params.SyncTimeout, false, params.SyncOnInitTimeout)),
 		// Local workloadmeta-backed tagger so the infraattributes processor can enrich
 		// spans with K8s tags (pod, namespace, deployment, ...) without a core agent

releasenotes/notes/otel-standalone-rc-configsync-fix-e11747b15bf4bc31.yaml

@@ -0,0 +1,16 @@
+# Each section from every release note are combined when the
+# CHANGELOG.rst is rendered. So the text needs to be worded so that
+# it does not depend on any information only available in another
+# section. This may mean repeating some details, but each section
+# must be readable independently of the other.
+#
+# Each section note must be formatted as reStructuredText.
+---
+fixes:
+  - |
+    Fixed a crash loop in standalone ``otel-agent`` (``DD_OTEL_STANDALONE=true``) when deployed
+    alongside a core Datadog Agent that injects ``DD_REMOTE_CONFIGURATION_ENABLED=true`` or
+    ``DD_AGENT_IPC_CONFIG_REFRESH_INTERVAL`` into its environment, such as when using the
+    Datadog Operator. Standalone mode now reliably disables Remote Configuration and config
+    sync regardless of these environment variables, since it has no core agent IPC endpoint
+    to use them with.