fix: return fatal on certain error codes during first stream cycle

Signed-off-by: Alexandra Oberaigner <[email protected]>

Author: alexandraoberaigner

providers/flagd/pkg/configuration.go

@@ -27,6 +27,8 @@ const (
 	defaultHost                         = "localhost"
 	defaultResolver                     = rpc
 	defaultGracePeriod                  = 5
+	defaultRetryBackoffMs               = 1000
+	defaultRetryBackoffMaxMs            = 120000
 	defaultFatalStatusCodes             = ""
 
 	rpc       ResolverType = "rpc"
@@ -69,6 +71,8 @@ type ProviderConfiguration struct {
 	CustomSyncProviderUri            string
 	GrpcDialOptionsOverride          []grpc.DialOption
 	RetryGracePeriod                 int
+	RetryBackoffMs					 int
+	RetryBackoffMaxMs				 int
 	FatalStatusCodes                 []string
 
 	log logr.Logger
@@ -212,12 +216,8 @@ func (cfg *ProviderConfiguration) updateFromEnvVar() {
 	if targetUri := os.Getenv(flagdTargetUriEnvironmentVariableName); targetUri != "" {
 		cfg.TargetUri = targetUri
 	}
-	if gracePeriod := os.Getenv(flagdGracePeriodVariableName); gracePeriod != "" {
-		if seconds, err := strconv.Atoi(gracePeriod); err == nil {
-			cfg.RetryGracePeriod = seconds
-			cfg.RetryGracePeriod = getIntFromEnvVarOrDefault(flagdGracePeriodVariableName, defaultGracePeriod, cfg.log)
-		}
-	}
+
+	cfg.RetryGracePeriod = getIntFromEnvVarOrDefault(flagdGracePeriodVariableName, defaultGracePeriod, cfg.log)
 
 	var fatalStatusCodes string
 	if envVal := os.Getenv(flagdFatalStatusCodesVariableName); envVal != "" {
@@ -431,6 +431,20 @@ func WithRetryGracePeriod(gracePeriod int) ProviderOption {
 	}
 }
 
+// WithRetryBackoffMs sets the initial backoff duration (in milliseconds) for retrying failed connections
+func WithRetryBackoffMs(retryBackoffMs int) ProviderOption {
+	return func(p *ProviderConfiguration) {
+		p.RetryBackoffMs = retryBackoffMs
+	}
+}
+
+// WithRetryBackoffMaxMs sets the maximum backoff duration (in milliseconds) for retrying failed connections
+func WithRetryBackoffMaxMs(retryBackoffMaxMs int) ProviderOption {
+	return func(p *ProviderConfiguration) {
+		p.RetryBackoffMaxMs = retryBackoffMaxMs
+	}
+}
+
 // WithFatalStatusCodes allows to set a list of gRPC status codes, which will cause streams to give up
 // and put the provider in a PROVIDER_FATAL state
 func WithFatalStatusCodes(fatalStatusCodes []string) ProviderOption {

providers/flagd/pkg/service/in_process/service.go

@@ -112,6 +112,8 @@ type Configuration struct {
 	GrpcDialOptionsOverride []googlegrpc.DialOption
 	CertificatePath         string
 	RetryGracePeriod        int
+	RetryBackOffMs          int
+	RetryBackOffMaxMs       int
 	FatalStatusCodes        []string
 }