Allow dynamic replication set of TSDB blocks on store-gateways

cmd/mimir/config-descriptor.json

@@ -11655,6 +11655,38 @@
           "fieldValue": null,
           "fieldDefaultValue": null
         },
+        {
+          "kind": "block",
+          "name": "dynamic_replication",
+          "required": false,
+          "desc": "",
+          "blockEntries": [
+            {
+              "kind": "field",
+              "name": "enabled",
+              "required": false,
+              "desc": "Use a higher number of replicas for recent blocks. Useful to spread query load more evenly at the cost of slightly higher disk usage.",
+              "fieldValue": null,
+              "fieldDefaultValue": false,
+              "fieldFlag": "store-gateway.dynamic-replication.enabled",
+              "fieldType": "boolean",
+              "fieldCategory": "experimental"
+            },
+            {
+              "kind": "field",
+              "name": "max_time_threshold",
+              "required": false,
+              "desc": "Threshold of the most recent sample in a block used to determine it is eligible for higher than default replication. If a block has samples within this amount of time, it is considered recent and will be owned by more replicas.",
+              "fieldValue": null,
+              "fieldDefaultValue": 90000000000000,
+              "fieldFlag": "store-gateway.dynamic-replication.max-time-threshold",
+              "fieldType": "duration",
+              "fieldCategory": "experimental"
+            }
+          ],
+          "fieldValue": null,
+          "fieldDefaultValue": null
+        },
         {
           "kind": "field",
           "name": "enabled_tenants",

cmd/mimir/help-all.txt.tmpl

@@ -3191,6 +3191,10 @@ Usage of ./cmd/mimir/mimir:
     	How long to wait between SIGTERM and shutdown. After receiving SIGTERM, Mimir will report not-ready status via /ready endpoint.
   -store-gateway.disabled-tenants comma-separated-list-of-strings
     	Comma separated list of tenants that cannot be loaded by the store-gateway. If specified, and the store-gateway would normally load a given tenant for (via -store-gateway.enabled-tenants or sharding), it will be ignored instead.
+  -store-gateway.dynamic-replication.enabled
+    	[experimental] Use a higher number of replicas for recent blocks. Useful to spread query load more evenly at the cost of slightly higher disk usage.
+  -store-gateway.dynamic-replication.max-time-threshold duration
+    	[experimental] Threshold of the most recent sample in a block used to determine it is eligible for higher than default replication. If a block has samples within this amount of time, it is considered recent and will be owned by more replicas. (default 25h0m0s)
   -store-gateway.enabled-tenants comma-separated-list-of-strings
     	Comma separated list of tenants that can be loaded by the store-gateway. If specified, only blocks for these tenants will be loaded by the store-gateway, otherwise all tenants can be loaded. Subject to sharding.
   -store-gateway.sharding-ring.auto-forget-enabled

development/mimir-microservices-mode/config/grafana-agent.yaml

@@ -23,10 +23,12 @@ prometheus:
                 - 'ruler-2:8023'
                 - 'compactor:8006'
                 - 'query-frontend:8007'
-                - 'store-gateway-1:8008'
-                - 'store-gateway-2:8009'
-                - 'query-scheduler:8011'
+                - 'query-scheduler:8008'
+                - 'store-gateway-1:8011'
+                - 'store-gateway-2:8012'
+                - 'store-gateway-3:8013'
                 - 'memcached-exporter:9150'
+                - 'continuous-test:8090'
                 - 'load-generator:9900'
               labels:
                 cluster: 'docker-compose'

development/mimir-microservices-mode/config/mimir.yaml

@@ -128,7 +128,7 @@ compactor:
 
 store_gateway:
   sharding_ring:
-    replication_factor: 1
+    replication_factor: 3
     heartbeat_period:   5s
     heartbeat_timeout:  15s
     wait_stability_min_duration: 0

development/mimir-microservices-mode/config/otel-collector.yaml

@@ -57,20 +57,20 @@ receivers:
                 cluster: 'docker-compose'
                 namespace: 'mimir-microservices-mode'
                 container: 'query-frontend'
-        - job_name: mimir-microservices-mode/store-gateway
+        - job_name: mimir-microservices-mode/query-scheduler
           static_configs:
-            - targets: ['store-gateway-1:8008', 'store-gateway-2:8009']
+            - targets: ['query-scheduler:8008']
               labels:
                 cluster: 'docker-compose'
                 namespace: 'mimir-microservices-mode'
-                container: 'store-gateway'
-        - job_name: mimir-microservices-mode/query-scheduler
+                container: 'query-scheduler'
+        - job_name: mimir-microservices-mode/store-gateway
           static_configs:
-            - targets: ['query-scheduler:8011']
+            - targets: ['store-gateway-1:8011', 'store-gateway-2:8012', 'store-gateway-3:8013']
               labels:
                 cluster: 'docker-compose'
                 namespace: 'mimir-microservices-mode'
-                container: 'query-scheduler'
+                container: 'store-gateway'
         - job_name: mimir-microservices-mode/memcached-exporter
           static_configs:
             - targets: ['memcached-exporter:9150']

development/mimir-microservices-mode/config/prometheus.yaml

@@ -17,10 +17,13 @@ scrape_configs:
           - 'ruler-2:8023'
           - 'compactor:8006'
           - 'query-frontend:8007'
-          - 'store-gateway-1:8008'
-          - 'store-gateway-2:8009'
-          - 'query-scheduler:8011'
+          - 'query-scheduler:8008'
+          - 'store-gateway-1:8011'
+          - 'store-gateway-2:8012'
+          - 'store-gateway-3:8013'
+          - 'memcached-exporter:9150'
           - 'continuous-test:8090'
+          - 'load-generator:9900'
         labels:
           cluster: 'docker-compose'
           namespace: 'mimir-microservices-mode'

development/mimir-microservices-mode/docker-compose.jsonnet

@@ -45,7 +45,7 @@ std.manifestYamlDoc({
     self.distributor +
     self.ingesters +
     self.read_components +  // Querier, Frontend and query-scheduler, if enabled.
-    self.store_gateways +
+    self.store_gateways(3) +
     self.compactor +
     self.rulers(2) +
     self.alertmanagers(3) +
@@ -113,7 +113,7 @@ std.manifestYamlDoc({
         httpPort: 8005,
         extraArguments:
           // Use of scheduler is activated by `-querier.scheduler-address` option and setting -querier.frontend-address option to nothing.
-          if $._config.use_query_scheduler then '-querier.scheduler-address=query-scheduler:9011 -querier.frontend-address=' else '',
+          if $._config.use_query_scheduler then '-querier.scheduler-address=query-scheduler:9008 -querier.frontend-address=' else '',
       }),
 
       'query-frontend': mimirService({
@@ -124,14 +124,14 @@ std.manifestYamlDoc({
         extraArguments:
           '-query-frontend.max-total-query-length=8760h' +
           // Use of scheduler is activated by `-query-frontend.scheduler-address` option.
-          (if $._config.use_query_scheduler then ' -query-frontend.scheduler-address=query-scheduler:9011' else ''),
+          (if $._config.use_query_scheduler then ' -query-frontend.scheduler-address=query-scheduler:9008' else ''),
       }),
     } + (
       if $._config.use_query_scheduler then {
         'query-scheduler': mimirService({
           name: 'query-scheduler',
           target: 'query-scheduler',
-          httpPort: 8011,
+          httpPort: 8008,
           extraArguments: '-query-frontend.max-total-query-length=8760h',
         }),
       } else {}
@@ -167,20 +167,14 @@ std.manifestYamlDoc({
     for id in std.range(1, count)
   },
 
-  store_gateways:: {
-    'store-gateway-1': mimirService({
-      name: 'store-gateway-1',
+  store_gateways(count):: {
+    ['store-gateway-%d' % id]: mimirService({
+      name: 'store-gateway-' + id,
       target: 'store-gateway',
-      httpPort: 8008,
-      jaegerApp: 'store-gateway-1',
-    }),
-
-    'store-gateway-2': mimirService({
-      name: 'store-gateway-2',
-      target: 'store-gateway',
-      httpPort: 8009,
-      jaegerApp: 'store-gateway-2',
-    }),
+      httpPort: 8010 + id,
+      jaegerApp: 'store-gateway-%d' % id,
+    })
+    for id in std.range(1, count)
   },
 
   continuous_test:: {