Add env var SPLUNK_CONFIG_YAML for storing configuration YAML

Author: bjsignalfx

cmd/otelcol/main.go

@@ -18,6 +18,7 @@
 package main
 
 import (
+	"bytes"
 	"fmt"
 	"log"
 	"os"
@@ -26,6 +27,7 @@ import (
 
 	"go.opentelemetry.io/collector/component"
 	"go.opentelemetry.io/collector/service"
+	"go.opentelemetry.io/collector/service/parserprovider"
 	"go.uber.org/zap"
 
 	"github.com/signalfx/splunk-otel-collector/internal/components"
@@ -37,6 +39,7 @@ import (
 const (
 	ballastEnvVarName         = "SPLUNK_BALLAST_SIZE_MIB"
 	configEnvVarName          = "SPLUNK_CONFIG"
+	configYamlEnvVarName      = "SPLUNK_CONFIG_YAML"
 	configServerEnabledEnvVar = "SPLUNK_DEBUG_CONFIG_SERVER"
 	memLimitMiBEnvVarName     = "SPLUNK_MEMORY_LIMIT_MIB"
 	memTotalEnvVarName        = "SPLUNK_MEMORY_TOTAL_MIB"
@@ -57,8 +60,7 @@ func main() {
 	// TODO: Use same format as the collector
 	log.SetFlags(log.Ldate | log.Ltime | log.Lshortfile)
 
-	args := os.Args[1:]
-	if !contains(args, "-h") && !contains(args, "--help") {
+	if !contains(os.Args[1:], "-h") && !contains(os.Args[1:], "--help") {
 		checkRuntimeParams()
 	}
 
@@ -77,6 +79,7 @@ func main() {
 	}
 
 	parserProvider := configprovider.NewConfigSourceParserProvider(
+		newBaseParserProvider(),
 		zap.NewNop(), // The service logger is not available yet, setting it to NoP.
 		info,
 		configsources.Get()...,
@@ -109,44 +112,27 @@ func contains(arr []string, str string) bool {
 
 // Get the value of a key in an array
 // Support key/value with and with an equal sign
-func getKeyValue(args []string, argName string) string {
-	val := ""
-	for i, arg := range args {
+func getKeyValue(args []string, arg string) string {
+	argEq := arg + "="
+	for i := range args {
 		switch {
-		case strings.HasPrefix(arg, argName+"="):
-			s := strings.Split(arg, "=")
-			val = s[1]
-		case arg == argName:
-			i++
-			val = args[i]
+		case strings.HasPrefix(args[i], argEq):
+			if s := strings.SplitN(args[i], "=", 2); len(s) == 2 {
+				return s[1]
+			}
+		case args[i] == arg && i < (len(args)-1):
+			return args[i+1]
 		}
 	}
-	return val
+	return ""
 }
 
 // Check runtime parameters
 // Runtime parameters take priority over environment variables
 // Config and ballast flags are checked
 // Config and all memory env vars are checked
 func checkRuntimeParams() {
-	args := os.Args[1:]
-	config := ""
-
-	// Check if config flag was passed and its value differs from config env var.
-	// If so, log that it will be used instead of env var value and set env var with that value.
-	// This allows users to set `--config` and have it take priority when running from most contexts.
-	cliConfig := getKeyValue(args, "--config")
-	if cliConfig != "" {
-		config = os.Getenv(configEnvVarName)
-		if config != "" && config != cliConfig {
-			log.Printf(
-				"Both %v and '--config' were specified. Overriding %q environment variable value with %q for this session",
-				configEnvVarName, config, cliConfig,
-			)
-		}
-		os.Setenv(configEnvVarName, cliConfig)
-	}
-	setConfig()
+	setValidConfigPath()
 
 	// Set default total memory
 	memTotalSizeMiB := defaultMemoryTotalMiB
@@ -169,66 +155,119 @@ func checkRuntimeParams() {
 	// Check if memory ballast flag was passed
 	// If so, ensure memory ballast env var is not set
 	// Then set memory ballast and limit properly
-	ballastSize := getKeyValue(args, "--mem-ballast-size-mib")
+	ballastSize := getKeyValue(os.Args[1:], "--mem-ballast-size-mib")
 	if ballastSize != "" {
-		config = os.Getenv(ballastEnvVarName)
-		if config != "" {
-			log.Fatalf("Both %v and '--config' were specified, but only one is allowed", ballastEnvVarName)
+		if os.Getenv(ballastEnvVarName) != "" {
+			log.Fatalf("Both %v and '--mem-ballast-size-mib' were specified, but only one is allowed", ballastEnvVarName)
 		}
 		os.Setenv(ballastEnvVarName, ballastSize)
 	}
 	setMemoryBallast(memTotalSizeMiB)
 	setMemoryLimit(memTotalSizeMiB)
 }
 
-// Validate and set the configuration
-func setConfig() {
-	// Check if the config is specified via the env var.
-	config := os.Getenv(configEnvVarName)
-	// If not attempt to use a default config; supports Docker and local
-	if config == "" {
-		_, err := os.Stat(defaultDockerSAPMConfig)
-		if err == nil {
-			config = defaultDockerSAPMConfig
+// Validate then set config file path environment variable to specified config file path flag.
+// Validate then set the config file path flag to the specified config file path environment variable if not specified.
+// Calls setValidDefaultConfigPath to validate and set the config file path flag, config file path environment variable
+// to the default config file path if the config file path flag, config file path environment variable and config YAML
+// environment variable are not specified.
+func setValidConfigPath() {
+	configPathFlag := getKeyValue(os.Args[1:], "--config")
+	configPathVar := os.Getenv(configEnvVarName)
+	configYaml := os.Getenv(configYamlEnvVarName)
+
+	if configPathFlag == "" && configPathVar == "" {
+		if configYaml == "" {
+			setValidDefaultConfigPath()
+		} else {
+			log.Printf("Using environment variable %s for configuration", configYamlEnvVarName)
 		}
-		_, err = os.Stat(defaultLocalSAPMConfig)
-		if err == nil {
-			config = defaultLocalSAPMConfig
+		return
+	}
+
+	if configPathFlag != "" {
+		validateConfigPath(configPathFlag)
+
+		if configPathVar != "" && configPathVar != configPathFlag {
+			log.Printf("Both %v and '--config' were specified. Overriding %q environment variable value with %q for this session", configEnvVarName, configPathVar, configPathFlag)
 		}
-		if config == "" {
-			log.Fatalf("Unable to find the default configuration file, ensure %s environment variable is set properly", configEnvVarName)
+		os.Setenv(configEnvVarName, configPathFlag)
+
+		if configYaml != "" {
+			log.Printf("Both %s and '--config' were specified. Using '--config' value %s for this session", configYamlEnvVarName, configPathFlag)
 		}
-	} else {
-		// Check if file exists.
-		_, err := os.Stat(config)
-		if err != nil {
-			log.Fatalf("Unable to find the configuration file (%s) ensure %s environment variable is set properly", config, configEnvVarName)
+
+		log.Printf("Set config to %v", configPathFlag)
+		return
+	}
+
+	if configPathVar != "" {
+		validateConfigPath(configPathVar)
+
+		if !contains(os.Args[1:], "--config") {
+			// Inject the command line flag that controls the configuration.
+			os.Args = append(os.Args, "--config="+configPathVar)
+		}
+
+		if configYaml != "" {
+			log.Printf("Both %s and %s were specified. Using %s environment variable value %s for this session", configEnvVarName, configYamlEnvVarName, configEnvVarName, configPathVar)
 		}
+
+		log.Printf("Set config to %v", configPathVar)
+	}
+}
+
+// Validate and set the config file path flag and config file path environment variable to the default config file path.
+// Only Docker and local configuration supported.
+func setValidDefaultConfigPath() {
+	var path string
+	var err error
+
+	if _, err = os.Stat(defaultDockerSAPMConfig); err == nil {
+		path = defaultDockerSAPMConfig
 	}
 
-	switch config {
+	if _, err = os.Stat(defaultLocalSAPMConfig); err == nil {
+		path = defaultLocalSAPMConfig
+	}
+
+	if path == "" {
+		log.Fatalf("Unable to find the default configuration file, ensure %s environment variable is set properly", configEnvVarName)
+	}
+
+	validateConfigPath(path)
+
+	os.Setenv(configEnvVarName, path)
+
+	if !contains(os.Args[1:], "--config") {
+		// Inject the command line flag that controls the configuration.
+		os.Args = append(os.Args, "--config="+path)
+	}
+
+	log.Printf("Set config to %v", path)
+}
+
+func validateConfigPath(path string) {
+	// Check if the config file exists.
+	if _, err := os.Stat(path); err != nil {
+		log.Fatalf("Unable to find the configuration file (%s) ensure %s environment variable is set properly", path, configEnvVarName)
+	}
+
+	// Check environment variables required by default configuration.
+	switch path {
 	case
 		defaultDockerSAPMConfig,
 		defaultDockerOTLPConfig,
 		defaultLocalSAPMConfig,
 		defaultLocalOTLPConfig:
-		// The following environment variables are required.
-		// If any are missing stop here.
 		requiredEnvVars := []string{realmEnvVarName, tokenEnvVarName}
 		for _, v := range requiredEnvVars {
 			if len(os.Getenv(v)) == 0 {
 				log.Printf("Usage: %s=12345 %s=us0 %s", tokenEnvVarName, realmEnvVarName, os.Args[0])
-				log.Fatalf("ERROR: Missing required environment variable %s with default config path %s", v, config)
+				log.Fatalf("ERROR: Missing required environment variable %s with default config path %s", v, path)
 			}
 		}
 	}
-
-	args := os.Args[1:]
-	if !contains(args, "--config") {
-		// Inject the command line flag that controls the configuration.
-		os.Args = append(os.Args, "--config="+config)
-	}
-	log.Printf("Set config to %v", config)
 }
 
 // Validate and set the memory ballast
@@ -287,6 +326,19 @@ func setMemoryLimit(memTotalSizeMiB int) {
 	log.Printf("Set memory limit to %d MiB", memLimit)
 }
 
+// Returns a ParserProvider that reads configuration YAML from an environment variable when applicable.
+func newBaseParserProvider() parserprovider.ParserProvider {
+	configPathFlag := getKeyValue(os.Args[1:], "--config")
+	configPathVar := os.Getenv(configEnvVarName)
+	configYaml := os.Getenv(configYamlEnvVarName)
+
+	if configPathFlag == "" && configPathVar == "" && configYaml != "" {
+		return parserprovider.NewInMemory(bytes.NewBufferString(configYaml))
+	}
+
+	return parserprovider.Default()
+}
+
 func runInteractive(params service.CollectorSettings) error {
 	app, err := service.New(params)
 	if err != nil {

cmd/otelcol/main_test.go

@@ -60,7 +60,7 @@ func TestGetKeyValue(t *testing.T) {
 func TestCheckRuntimeParams(*testing.T) {
 	oldArgs := os.Args
 	os.Setenv(configEnvVarName, "../../"+defaultLocalSAPMConfig)
-	setConfig()
+	setValidConfigPath()
 	os.Unsetenv(configEnvVarName)
 	checkRuntimeParams()
 
@@ -101,7 +101,7 @@ func TestUseConfigFromEnvVar(t *testing.T) {
 	os.Setenv(tokenEnvVarName, "12345")
 	os.Setenv(realmEnvVarName, "us0")
 	os.Setenv(configEnvVarName, "../../"+defaultLocalSAPMConfig)
-	setConfig()
+	setValidConfigPath()
 
 	args := os.Args[1:]
 	c := getKeyValue(args, "--config")

docs/getting-started/linux-manual.md

@@ -139,13 +139,21 @@ $ docker run --rm -e SPLUNK_ACCESS_TOKEN=12345 -e SPLUNK_REALM=us0 \
 
 ### Custom Configuration
 
-When changes to the default configuration file is needed, a custom
-configuration file can specified and used instead. Use the `SPLUNK_CONFIG`
-environment variable or the `--config` command line argument to provide a
-custom configuration.
+When changes to the default configuration YAML file are needed, create a
+custom configuration file. Use environment variable `SPLUNK_CONFIG` or
+command line argument `--config` to provide the path to this file.
+
+Also, you can use environment variable `SPLUNK_CONFIG_YAML` to specify 
+your custom configuration YAML at the command line. This is useful in
+environments where access to the underlying file system is not readily
+available. For example, in AWS Fargate you can store your custom configuration
+YAML in a parameter in AWS Systems Manager Parameter Store, then in
+your container definition specify `SPLUNK_CONFIG_YAML` to get the
+configuration from the parameter.
 
 > Command line arguments take precedence over environment variables. This
-> applies to `--config` and `--mem-ballast-size-mib`.
+> applies to `--config` and `--mem-ballast-size-mib`. `SPLUNK_CONFIG` 
+> takes precedence over `SPLUNK_CONFIG_YAML`
 
 For example in Docker:
 
@@ -160,11 +168,41 @@ $ docker run --rm -e SPLUNK_ACCESS_TOKEN=12345 -e SPLUNK_REALM=us0 \
 
 > Use of a SemVer tag over `latest` is highly recommended.
 
-In the case of Docker, a volume mount may be required to load custom
-configuration as shown above.
+In the case of Docker, a volume mount may be required to load the custom
+configuration file, as shown above.
 
-If the custom configuration includes a `memory_limiter` processor then the
+If the custom configuration includes a `memory_limiter` processor, then the
 `ballast_size_mib` parameter should be the same as the
 `SPLUNK_BALLAST_SIZE_MIB` environment variable. See
 [gateway_config.yaml](../../cmd/otelcol/config/collector/gateway_config.yaml)
 as an example.
+
+The following example shows the `SPLUNK_CONFIG_YAML` environment variable:
+```bash
+$ CONFIG_YAML=$(cat <<-END
+receivers:
+   hostmetrics:
+      collection_interval: 1s
+      scrapers:
+         cpu:
+exporters:
+   logging:
+      logLevel: debug
+service:
+   pipelines:
+      metrics:
+         receivers: [hostmetrics]
+         exporters: [logging]
+END
+)
+
+$ docker run --rm \
+    -e SPLUNK_CONFIG_YAML=${CONFIG_YAML} \
+    --name otelcol quay.io/signalfx/splunk-otel-collector:latest
+```
+The configuration YAML above is for collecting and logging cpu
+metrics. The YAML is assigned to parameter `CONFIG_YAML` for
+convenience in the first command. In the subsequent `docker run`
+command, parameter `CONFIG_YAML` is expanded and assigned to
+environment variable `SPLUNK_CONFIG_YAML`. Note that YAML 
+requires whitespace indentation to be maintained.

internal/configprovider/config_source_provider.go

@@ -37,9 +37,12 @@ type configSourceParserProvider struct {
 }
 
 // NewConfigSourceParserProvider creates a ParserProvider that uses config sources.
-func NewConfigSourceParserProvider(logger *zap.Logger, buildInfo component.BuildInfo, factories ...Factory) parserprovider.ParserProvider {
+func NewConfigSourceParserProvider(pp parserprovider.ParserProvider, logger *zap.Logger, buildInfo component.BuildInfo, factories ...Factory) parserprovider.ParserProvider {
+	if pp == nil {
+		pp = parserprovider.Default()
+	}
 	return &configSourceParserProvider{
-		pp:        parserprovider.Default(),
+		pp:        pp,
 		logger:    logger,
 		factories: factories,
 		buildInfo: buildInfo,

internal/configprovider/config_source_provider_test.go

@@ -88,6 +88,7 @@ func TestConfigSourceParserProvider(t *testing.T) {
 			}
 
 			pp := NewConfigSourceParserProvider(
+				parserprovider.Default(),
 				zap.NewNop(),
 				component.DefaultBuildInfo(),
 				factories...,