feat(runtime-config): Add default tenant ID for runtime config fallback

Add -runtime-config.default-tenant-id flag that enables a synthetic tenant
entry in the runtime config overrides to serve as the default for all tenants
without per-tenant overrides. This allows changing default limits via the
config pipeline (hot-reloaded) instead of requiring a Cortex restart.

Resolution order: per-tenant override → default tenant → CLI flags.

When the flag is empty (default), the feature is disabled and behavior is
unchanged.

This is an experimental feature.

Signed-off-by: Ben Ye <benye@amazon.com>

Author: yeya24

docs/configuration/config-file-reference.md

@@ -6382,6 +6382,12 @@ filesystem:
   # Local filesystem storage directory.
   # CLI flag: -runtime-config.filesystem.dir
   [dir: <string> | default = ""]
+
+# [Experimental] Synthetic tenant ID used as fallback for runtime config
+# defaults. When set, overrides for this tenant ID apply to all tenants without
+# per-tenant overrides, before falling back to CLI flag defaults.
+# CLI flag: -runtime-config.default-tenant-id
+[default_tenant_id: <string> | default = ""]
 ```
 
 ### `s3_sse_config`

docs/configuration/v1-guarantees.md

@@ -133,3 +133,5 @@ Currently experimental features are:
 - Ingester: Active Series Tracker
   - Per-tenant `active_series_trackers` configuration in runtime config overrides
   - Counts active series matching PromQL label matchers and exposes `cortex_ingester_active_series_per_tracker` metric
+- Runtime Config: Default Tenant ID
+  - `-runtime-config.default-tenant-id` (string) - Synthetic tenant ID used as fallback for runtime config defaults. Overrides for this tenant apply to all tenants without per-tenant overrides.

pkg/cortex/modules.go

@@ -206,7 +206,7 @@ func (t *Cortex) initRuntimeConfig() (services.Service, error) {
 }
 
 func (t *Cortex) initOverridesConfig() (services.Service, error) {
-	t.OverridesConfig = validation.NewOverrides(t.Cfg.LimitsConfig, t.TenantLimits)
+	t.OverridesConfig = validation.NewOverridesWithDefaultTenantID(t.Cfg.LimitsConfig, t.TenantLimits, t.Cfg.RuntimeConfig.DefaultTenantID)
 	// overrides don't have operational state, nor do they need to do anything more in starting/stopping phase,
 	// so there is no need to return any service.
 	return nil, nil

pkg/util/runtimeconfig/manager.go

@@ -36,12 +36,19 @@ type Config struct {
 	Loader   Loader `yaml:"-"`
 
 	StorageConfig bucket.Config `yaml:",inline"`
+
+	// DefaultTenantID is the synthetic tenant ID used as a fallback for default
+	// runtime config values. When set, overrides for this tenant ID are applied
+	// to all tenants that do not have their own per-tenant override, before
+	// falling back to CLI flag defaults. This is an experimental feature.
+	DefaultTenantID string `yaml:"default_tenant_id"`
 }
 
 // RegisterFlags registers flags.
 func (mc *Config) RegisterFlags(f *flag.FlagSet) {
 	f.StringVar(&mc.LoadPath, "runtime-config.file", "", "File with the configuration that can be updated in runtime.")
 	f.DurationVar(&mc.ReloadPeriod, "runtime-config.reload-period", 10*time.Second, "How often to check runtime config file.")
+	f.StringVar(&mc.DefaultTenantID, "runtime-config.default-tenant-id", "", "[Experimental] Synthetic tenant ID used as fallback for runtime config defaults. When set, overrides for this tenant ID apply to all tenants without per-tenant overrides, before falling back to CLI flag defaults.")
 
 	mc.StorageConfig.RegisterFlagsWithPrefixAndBackend("runtime-config.", f, bucket.Filesystem)
 }

pkg/util/validation/limits.go

@@ -690,8 +690,9 @@ type TenantLimits interface {
 // Overrides periodically fetch a set of per-user overrides, and provides convenience
 // functions for fetching the correct value.
 type Overrides struct {
-	defaultLimits *Limits
-	tenantLimits  TenantLimits
+	defaultLimits   *Limits
+	tenantLimits    TenantLimits
+	defaultTenantID string
 }
 
 // NewOverrides makes a new Overrides.
@@ -702,6 +703,18 @@ func NewOverrides(defaults Limits, tenantLimits TenantLimits) *Overrides {
 	}
 }
 
+// NewOverridesWithDefaultTenantID creates Overrides with a synthetic default tenant ID.
+// When a tenant has no per-tenant override and defaultTenantID is non-empty,
+// the overrides for defaultTenantID are used before falling back to CLI flag defaults.
+// This is an experimental feature.
+func NewOverridesWithDefaultTenantID(defaults Limits, tenantLimits TenantLimits, defaultTenantID string) *Overrides {
+	return &Overrides{
+		tenantLimits:    tenantLimits,
+		defaultLimits:   &defaults,
+		defaultTenantID: defaultTenantID,
+	}
+}
+
 // IngestionRate returns the limit on ingester rate (samples per second).
 func (o *Overrides) IngestionRate(userID string) float64 {
 	return o.GetOverridesForUser(userID).IngestionRate
@@ -1288,6 +1301,12 @@ func (o *Overrides) GetOverridesForUser(userID string) *Limits {
 		if l != nil {
 			return l
 		}
+		if o.defaultTenantID != "" && userID != o.defaultTenantID {
+			l = o.tenantLimits.ByUserID(o.defaultTenantID)
+			if l != nil {
+				return l
+			}
+		}
 	}
 	return o.defaultLimits
 }

pkg/util/validation/limits_test.go

@@ -266,6 +266,82 @@ func TestOverridesManager_GetOverrides(t *testing.T) {
 	require.Equal(t, 0, ov.MaxLabelsSizeBytes("user2"))
 }
 
+func TestOverrides_DefaultTenantFallback(t *testing.T) {
+	defaultTenantID := "__default__"
+
+	tenantLimits := map[string]*Limits{}
+
+	cliDefaults := Limits{
+		IngestionRate:          10000,
+		MaxGlobalSeriesPerUser: 100000,
+		MaxLabelNamesPerSeries: 30,
+	}
+
+	// Set up default tenant override with higher ingestion rate
+	defaultTenantOverride := cliDefaults
+	defaultTenantOverride.IngestionRate = 50000
+	defaultTenantOverride.MaxGlobalSeriesPerUser = 500000
+	tenantLimits[defaultTenantID] = &defaultTenantOverride
+
+	ov := NewOverridesWithDefaultTenantID(cliDefaults, newMockTenantLimits(tenantLimits), defaultTenantID)
+
+	// Tenant without override should get default tenant values
+	require.Equal(t, float64(50000), ov.IngestionRate("user1"))
+	require.Equal(t, 500000, ov.MaxGlobalSeriesPerUser("user1"))
+	require.Equal(t, 30, ov.MaxLabelNamesPerSeries("user1"))
+
+	// Add a per-tenant override for user2
+	user2Limits := cliDefaults
+	user2Limits.IngestionRate = 200000
+	tenantLimits["user2"] = &user2Limits
+
+	// user2 should get its own override, not the default tenant
+	require.Equal(t, float64(200000), ov.IngestionRate("user2"))
+	// user2's MaxGlobalSeriesPerUser comes from its own override (which copied cliDefaults)
+	require.Equal(t, 100000, ov.MaxGlobalSeriesPerUser("user2"))
+
+	// The default tenant itself should get its own limits (not recurse)
+	require.Equal(t, float64(50000), ov.IngestionRate(defaultTenantID))
+	require.Equal(t, 500000, ov.MaxGlobalSeriesPerUser(defaultTenantID))
+}
+
+func TestOverrides_DefaultTenantDisabled(t *testing.T) {
+	tenantLimits := map[string]*Limits{}
+
+	cliDefaults := Limits{
+		IngestionRate:          10000,
+		MaxGlobalSeriesPerUser: 100000,
+	}
+
+	// Default tenant override exists in the map but feature is disabled (empty ID)
+	defaultOverride := cliDefaults
+	defaultOverride.IngestionRate = 50000
+	tenantLimits["__default__"] = &defaultOverride
+
+	ov := NewOverridesWithDefaultTenantID(cliDefaults, newMockTenantLimits(tenantLimits), "")
+
+	// With empty defaultTenantID, feature is disabled — should get CLI defaults
+	require.Equal(t, float64(10000), ov.IngestionRate("user1"))
+	require.Equal(t, 100000, ov.MaxGlobalSeriesPerUser("user1"))
+}
+
+func TestOverrides_DefaultTenantNotInMap(t *testing.T) {
+	tenantLimits := map[string]*Limits{}
+
+	cliDefaults := Limits{
+		IngestionRate:          10000,
+		MaxGlobalSeriesPerUser: 100000,
+	}
+
+	// Feature enabled but no __default__ entry in runtime config
+	ov := NewOverridesWithDefaultTenantID(cliDefaults, newMockTenantLimits(tenantLimits), "__default__")
+
+	// Should fall through to CLI defaults
+	require.Equal(t, float64(10000), ov.IngestionRate("user1"))
+	require.Equal(t, 100000, ov.MaxGlobalSeriesPerUser("user1"))
+}
+
+
 func TestLimitsLoadingFromYaml(t *testing.T) {
 	SetDefaultLimitsForYAMLUnmarshalling(Limits{
 		MaxLabelNameLength: 100,

schemas/cortex-config-schema.json

@@ -7810,6 +7810,11 @@
           "type": "string",
           "x-cli-flag": "runtime-config.backend"
         },
+        "default_tenant_id": {
+          "description": "[Experimental] Synthetic tenant ID used as fallback for runtime config defaults. When set, overrides for this tenant ID apply to all tenants without per-tenant overrides, before falling back to CLI flag defaults.",
+          "type": "string",
+          "x-cli-flag": "runtime-config.default-tenant-id"
+        },
         "file": {
           "description": "File with the configuration that can be updated in runtime.",
           "type": "string",