diff --git a/CHANGELOG.md b/CHANGELOG.md
index 3bbdac43d65..d8a0a9c04b8 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -15,6 +15,7 @@
 * [CHANGE] Histogram `cortex_memcache_request_duration_seconds` `method` label value changes from `Memcached.Get` to `Memcached.GetBatched` for batched lookups, and is not reported for non-batched lookups (label value `Memcached.GetMulti` remains, and had exactly the same value as `Get` in nonbatched lookups).  The same change applies to tracing spans. #3046
 * [CHANGE] TLS server validation is now enabled by default, a new parameter `tls_insecure_skip_verify` can be set to true to skip validation optionally. #3030
 * [CHANGE] `cortex_ruler_config_update_failures_total` has been removed in favor of `cortex_ruler_config_last_reload_successful`. #3056
+* [CHANGE] `ruler.evaluation_delay_duration` field in YAML config has been moved and renamed to `limits.ruler_evaluation_delay_duration`. #3098
 * [CHANGE] Removed obsolete `results_cache.max_freshness` from YAML config (deprecated since Cortex 1.2). #3145
 * [CHANGE] Removed obsolete `-promql.lookback-delta` option (deprecated since Cortex 1.2, replaced with `-querier.lookback-delta`). #3144
 * [FEATURE] Logging of the source IP passed along by a reverse proxy is now supported by setting the `-server.log-source-ips-enabled`. For non standard headers the settings `-server.log-source-ips-header` and `-server.log-source-ips-regex` can be used. #2985
@@ -31,6 +32,7 @@
 * [ENHANCEMENT] Add "integration" as a label for `cortex_alertmanager_notifications_total` and `cortex_alertmanager_notifications_failed_total` metrics. #3056
 * [ENHANCEMENT] Add `cortex_ruler_config_last_reload_successful` and `cortex_ruler_config_last_reload_successful_seconds` to check status of users rule manager. #3056
 * [ENHANCEMENT] Memcached dial() calls now have an optional circuit-breaker to avoid hammering a broken cache #3051
+* [ENHANCEMENT] `-ruler.evaluation-delay-duration` is now overridable as a per-tenant limit, `ruler_evaluation_delay_duration`. #3098
 * [ENHANCEMENT] Add TLS support to etcd client. #3102
 * [ENHANCEMENT] When a tenant accesses the Alertmanager UI or its API, if we have valid `-alertmanager.configs.fallback` we'll use that to start the manager and avoid failing the request. #3073
 * [ENHANCEMENT] Add `DELETE api/v1/rules/{namespace}` to the Ruler. It allows all the rule groups of a namespace to be deleted. #3120
diff --git a/docs/configuration/config-file-reference.md b/docs/configuration/config-file-reference.md
index 4a46d535319..4007c184472 100644
--- a/docs/configuration/config-file-reference.md
+++ b/docs/configuration/config-file-reference.md
@@ -846,9 +846,8 @@ ruler_client:
 # CLI flag: -ruler.evaluation-interval
 [evaluation_interval: <duration> | default = 1m]
 
-# Duration to delay the evaluation of rules to ensure they underlying metrics
-# have been pushed to cortex.
-# CLI flag: -ruler.evaluation-delay-duration
+# Deprecated. Please use -ruler.evaluation-delay-duration instead.
+# CLI flag: -ruler.evaluation-delay-duration-deprecated
 [evaluation_delay_duration: <duration> | default = 0s]
 
 # How frequently to poll for rule changes
@@ -2822,6 +2821,11 @@ The `limits_config` configures default and per-tenant limits imposed by Cortex s
 # CLI flag: -frontend.max-cache-freshness
 [max_cache_freshness: <duration> | default = 1m]
 
+# Duration to delay the evaluation of rules to ensure the underlying metrics
+# have been pushed to Cortex.
+# CLI flag: -ruler.evaluation-delay-duration
+[ruler_evaluation_delay_duration: <duration> | default = 0s]
+
 # The default tenant's shard size when the shuffle-sharding strategy is used.
 # Must be set when the store-gateway sharding is enabled with the
 # shuffle-sharding strategy. When this setting is specified in the per-tenant
diff --git a/pkg/cortex/modules.go b/pkg/cortex/modules.go
index 29bc5a70ced..604660ad6ea 100644
--- a/pkg/cortex/modules.go
+++ b/pkg/cortex/modules.go
@@ -35,6 +35,7 @@ import (
 	"github.com/cortexproject/cortex/pkg/ruler"
 	"github.com/cortexproject/cortex/pkg/storegateway"
 	"github.com/cortexproject/cortex/pkg/util"
+	"github.com/cortexproject/cortex/pkg/util/flagext"
 	"github.com/cortexproject/cortex/pkg/util/modules"
 	"github.com/cortexproject/cortex/pkg/util/runtimeconfig"
 	"github.com/cortexproject/cortex/pkg/util/services"
@@ -126,6 +127,19 @@ func (t *Cortex) initRing() (serv services.Service, err error) {
 }
 
 func (t *Cortex) initRuntimeConfig() (services.Service, error) {
+	// We need to modify LimitsConfig before calling SetDefaultLimitsForYAMLUnmarshalling later in this method
+	// but also if runtime-config is not used, for setting limits used by initOverrides.
+	// TODO: Remove this in Cortex 1.6.
+	if t.Cfg.Ruler.EvaluationDelay != 0 && t.Cfg.LimitsConfig.RulerEvaluationDelay == 0 {
+		t.Cfg.LimitsConfig.RulerEvaluationDelay = t.Cfg.Ruler.EvaluationDelay
+
+		// No need to report if this field isn't going to be used.
+		if t.Cfg.Target == All || t.Cfg.Target == Ruler {
+			flagext.DeprecatedFlagsUsed.Inc()
+			level.Warn(util.Logger).Log("msg", "Using DEPRECATED YAML config field ruler.evaluation_delay_duration, please use limits.ruler_evaluation_delay_duration instead.")
+		}
+	}
+
 	if t.Cfg.RuntimeConfig.LoadPath == "" {
 		t.Cfg.RuntimeConfig.LoadPath = t.Cfg.LimitsConfig.PerTenantOverrideConfig
 		t.Cfg.RuntimeConfig.ReloadPeriod = t.Cfg.LimitsConfig.PerTenantOverridePeriod
@@ -514,7 +528,7 @@ func (t *Cortex) initRuler() (serv services.Service, err error) {
 	rulerRegisterer := prometheus.WrapRegistererWith(prometheus.Labels{"engine": "ruler"}, prometheus.DefaultRegisterer)
 	queryable, engine := querier.New(t.Cfg.Querier, t.Overrides, t.Distributor, t.StoreQueryables, t.TombstonesLoader, rulerRegisterer)
 
-	managerFactory := ruler.DefaultTenantManagerFactory(t.Cfg.Ruler, t.Distributor, queryable, engine)
+	managerFactory := ruler.DefaultTenantManagerFactory(t.Cfg.Ruler, t.Distributor, queryable, engine, t.Overrides)
 	manager, err := ruler.NewDefaultMultiTenantManager(t.Cfg.Ruler, managerFactory, prometheus.DefaultRegisterer, util.Logger)
 	if err != nil {
 		return nil, err
diff --git a/pkg/ruler/compat.go b/pkg/ruler/compat.go
index ff0c424fdd0..494e81b5de0 100644
--- a/pkg/ruler/compat.go
+++ b/pkg/ruler/compat.go
@@ -72,12 +72,21 @@ func (t *PusherAppendable) Appender(ctx context.Context) storage.Appender {
 	}
 }
 
+// RulesLimits is the one function we need from limits.Overrides, and
+// is here to limit coupling.
+type RulesLimits interface {
+	EvaluationDelay(usedID string) time.Duration
+}
+
 // engineQueryFunc returns a new query function using the rules.EngineQueryFunc function
 // and passing an altered timestamp.
-func engineQueryFunc(engine *promql.Engine, q storage.Queryable, delay time.Duration) rules.QueryFunc {
-	orig := rules.EngineQueryFunc(engine, q)
+func engineQueryFunc(engine *promql.Engine, q storage.Queryable, overrides RulesLimits, userID string) rules.QueryFunc {
 	return func(ctx context.Context, qs string, t time.Time) (promql.Vector, error) {
-		return orig(ctx, qs, t.Add(-delay))
+		orig := rules.EngineQueryFunc(engine, q)
+		// Delay the evaluation of all rules by a set interval to give a buffer
+		// to metric that haven't been forwarded to cortex yet.
+		evaluationDelay := overrides.EvaluationDelay(userID)
+		return orig(ctx, qs, t.Add(-evaluationDelay))
 	}
 }
 
@@ -94,6 +103,7 @@ func DefaultTenantManagerFactory(
 	p Pusher,
 	q storage.Queryable,
 	engine *promql.Engine,
+	overrides RulesLimits,
 ) ManagerFactory {
 	return func(
 		ctx context.Context,
@@ -105,7 +115,7 @@ func DefaultTenantManagerFactory(
 		return rules.NewManager(&rules.ManagerOptions{
 			Appendable:      &PusherAppendable{pusher: p, userID: userID},
 			Queryable:       q,
-			QueryFunc:       engineQueryFunc(engine, q, cfg.EvaluationDelay),
+			QueryFunc:       engineQueryFunc(engine, q, overrides, userID),
 			Context:         user.InjectOrgID(ctx, userID),
 			ExternalURL:     cfg.ExternalURL.URL,
 			NotifyFunc:      SendAlerts(notifier, cfg.ExternalURL.URL.String()),
diff --git a/pkg/ruler/ruler.go b/pkg/ruler/ruler.go
index 2e68b722897..7adce6bf2d1 100644
--- a/pkg/ruler/ruler.go
+++ b/pkg/ruler/ruler.go
@@ -49,8 +49,8 @@ type Config struct {
 	ClientTLSConfig tls.ClientConfig `yaml:"ruler_client"`
 	// How frequently to evaluate rules by default.
 	EvaluationInterval time.Duration `yaml:"evaluation_interval"`
-	// Delay the evaluation of all rules by a set interval to give a buffer
-	// to metric that haven't been forwarded to cortex yet.
+	// Deprecated. Replaced with pkg/util/validation/Limits.RulerEvaluationDelay field.
+	// TODO: To be removed in Cortex 1.6.
 	EvaluationDelay time.Duration `yaml:"evaluation_delay_duration"`
 	// How frequently to poll for updated rules.
 	PollInterval time.Duration `yaml:"poll_interval"`
@@ -110,7 +110,7 @@ func (cfg *Config) RegisterFlags(f *flag.FlagSet) {
 	cfg.ExternalURL.URL, _ = url.Parse("") // Must be non-nil
 	f.Var(&cfg.ExternalURL, "ruler.external.url", "URL of alerts return path.")
 	f.DurationVar(&cfg.EvaluationInterval, "ruler.evaluation-interval", 1*time.Minute, "How frequently to evaluate rules")
-	f.DurationVar(&cfg.EvaluationDelay, "ruler.evaluation-delay-duration", 0, "Duration to delay the evaluation of rules to ensure they underlying metrics have been pushed to cortex.")
+	f.DurationVar(&cfg.EvaluationDelay, "ruler.evaluation-delay-duration-deprecated", 0, "Deprecated. Please use -ruler.evaluation-delay-duration instead.")
 	f.DurationVar(&cfg.PollInterval, "ruler.poll-interval", 1*time.Minute, "How frequently to poll for rule changes")
 
 	f.StringVar(&cfg.AlertmanagerURL, "ruler.alertmanager-url", "", "Comma-separated list of URL(s) of the Alertmanager(s) to send notifications to. Each Alertmanager URL is treated as a separate group in the configuration. Multiple Alertmanagers in HA per group can be supported by using DNS resolution via -ruler.alertmanager-discovery.")
diff --git a/pkg/ruler/ruler_test.go b/pkg/ruler/ruler_test.go
index 046d6c5d5f0..c2abb714c08 100644
--- a/pkg/ruler/ruler_test.go
+++ b/pkg/ruler/ruler_test.go
@@ -57,7 +57,13 @@ func defaultRulerConfig(store rules.RuleStore) (Config, func()) {
 	return cfg, cleanup
 }
 
-func testSetup(t *testing.T, cfg Config) (*promql.Engine, storage.QueryableFunc, Pusher, log.Logger, func()) {
+type ruleLimits time.Duration
+
+func (r ruleLimits) EvaluationDelay(_ string) time.Duration {
+	return time.Duration(r)
+}
+
+func testSetup(t *testing.T, cfg Config) (*promql.Engine, storage.QueryableFunc, Pusher, log.Logger, RulesLimits, func()) {
 	dir, err := ioutil.TempDir("", t.Name())
 	testutil.Ok(t, err)
 	cleanup := func() {
@@ -83,24 +89,24 @@ func testSetup(t *testing.T, cfg Config) (*promql.Engine, storage.QueryableFunc,
 	l := log.NewLogfmtLogger(os.Stdout)
 	l = level.NewFilter(l, level.AllowInfo())
 
-	return engine, noopQueryable, pusher, l, cleanup
+	return engine, noopQueryable, pusher, l, ruleLimits(0), cleanup
 }
 
 func newManager(t *testing.T, cfg Config) (*DefaultMultiTenantManager, func()) {
-	engine, noopQueryable, pusher, logger, cleanup := testSetup(t, cfg)
-	manager, err := NewDefaultMultiTenantManager(cfg, DefaultTenantManagerFactory(cfg, pusher, noopQueryable, engine), prometheus.NewRegistry(), logger)
+	engine, noopQueryable, pusher, logger, overrides, cleanup := testSetup(t, cfg)
+	manager, err := NewDefaultMultiTenantManager(cfg, DefaultTenantManagerFactory(cfg, pusher, noopQueryable, engine, overrides), prometheus.NewRegistry(), logger)
 	require.NoError(t, err)
 
 	return manager, cleanup
 }
 
 func newRuler(t *testing.T, cfg Config) (*Ruler, func()) {
-	engine, noopQueryable, pusher, logger, cleanup := testSetup(t, cfg)
+	engine, noopQueryable, pusher, logger, overrides, cleanup := testSetup(t, cfg)
 	storage, err := NewRuleStorage(cfg.StoreConfig)
 	require.NoError(t, err)
 
 	reg := prometheus.NewRegistry()
-	managerFactory := DefaultTenantManagerFactory(cfg, pusher, noopQueryable, engine)
+	managerFactory := DefaultTenantManagerFactory(cfg, pusher, noopQueryable, engine, overrides)
 	manager, err := NewDefaultMultiTenantManager(cfg, managerFactory, reg, util.Logger)
 	require.NoError(t, err)
 
diff --git a/pkg/util/validation/limits.go b/pkg/util/validation/limits.go
index e88bb47b2b8..3d4a7ba80fb 100644
--- a/pkg/util/validation/limits.go
+++ b/pkg/util/validation/limits.go
@@ -62,6 +62,9 @@ type Limits struct {
 	CardinalityLimit    int           `yaml:"cardinality_limit"`
 	MaxCacheFreshness   time.Duration `yaml:"max_cache_freshness"`
 
+	// Ruler defaults and limits.
+	RulerEvaluationDelay time.Duration `yaml:"ruler_evaluation_delay_duration"`
+
 	// Store-gateway.
 	StoreGatewayTenantShardSize int `yaml:"store_gateway_tenant_shard_size"`
 
@@ -109,6 +112,8 @@ func (l *Limits) RegisterFlags(f *flag.FlagSet) {
 	f.IntVar(&l.CardinalityLimit, "store.cardinality-limit", 1e5, "Cardinality limit for index queries. This limit is ignored when running the Cortex blocks storage. 0 to disable.")
 	f.DurationVar(&l.MaxCacheFreshness, "frontend.max-cache-freshness", 1*time.Minute, "Most recent allowed cacheable result per-tenant, to prevent caching very recent results that might still be in flux.")
 
+	f.DurationVar(&l.RulerEvaluationDelay, "ruler.evaluation-delay-duration", 0, "Duration to delay the evaluation of rules to ensure the underlying metrics have been pushed to Cortex.")
+
 	f.StringVar(&l.PerTenantOverrideConfig, "limits.per-user-override-config", "", "File name of per-user overrides. [deprecated, use -runtime-config.file instead]")
 	f.DurationVar(&l.PerTenantOverridePeriod, "limits.per-user-override-period", 10*time.Second, "Period with which to reload the overrides. [deprecated, use -runtime-config.reload-period instead]")
 
@@ -346,6 +351,11 @@ func (o *Overrides) SubringSize(userID string) int {
 	return o.getOverridesForUser(userID).SubringSize
 }
 
+// EvaluationDelay returns the rules evaluation delay for a given user.
+func (o *Overrides) EvaluationDelay(userID string) time.Duration {
+	return o.getOverridesForUser(userID).RulerEvaluationDelay
+}
+
 // StoreGatewayTenantShardSize returns the size of the store-gateway shard size for a given user.
 func (o *Overrides) StoreGatewayTenantShardSize(userID string) int {
 	return o.getOverridesForUser(userID).StoreGatewayTenantShardSize