grafana · narqo · Aug 12, 2026 · Jul 11, 2026 · Jul 11, 2026 · Aug 6, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -11,6 +11,7 @@
 * [ENHANCEMENT] Distributor: Add the experimental `cortex_distributor_otlp_requests_with_job_or_instance_resource_attribute_total{user}` counter to track OTLP requests carrying `job` or `instance` as a resource attribute. #16285
 * [ENHANCEMENT] Ruler: Split oversized remote distributor writes to keep resulting calls within the configured gRPC maximum send size, and expose the number of generated requests in `cortex_ruler_remote_distributor_requests_per_write_request`. #16160
 * [ENHANCEMENT] Alerts: Don't fire `MimirMemberlistZoneAwareRoutingAutoFailover` while the node is still joining the cluster. #16315
+* [FEATURE] Querier: Add experimental per-tenant limit `-querier.max-blocks-per-store-request` to cap the number of blocks a single store-gateway request may reference. Disabled by default. #16292
 * [BUGFIX] Query-frontend: Wait for the querier ring to be populated during startup, up to 30 seconds, before reporting the query-frontend as ready. Previously a query-frontend could become ready before it had seen any querier in the ring and fail every query it received until the ring was populated. Only applies when remote execution is enabled, and can be disabled with the experimental `-query-frontend.wait-for-querier-ring-on-startup=false`. #16333
 * [BUGFIX] Query-frontend: Fail queries with a clear error, rather than planning them against an invalid maximum supported query plan version, when the querier ring contains only unhealthy queriers. #16333
 * [BUGFIX] Query-frontend: Return a HTTP 500 error rather than a HTTP 400 when a querier receives a query plan that is too new. #16233

@@ -6085,6 +6085,17 @@
           "fieldType": "int",
           "fieldCategory": "experimental"
         },
+        {
+          "kind": "field",
+          "name": "max_blocks_per_store_request",
+          "required": false,
+          "desc": "Maximum number of blocks that a querier will reference in a single request to a store-gateway. When a request would exceed this, it is split into multiple requests to the same store-gateway. 0 disables the limit.",
+          "fieldValue": null,
+          "fieldDefaultValue": 0,
+          "fieldFlag": "querier.max-blocks-per-store-request",
+          "fieldType": "int",
+          "fieldCategory": "experimental"
+        },
         {
           "kind": "field",
           "name": "max_query_lookback",

@@ -2473,6 +2473,8 @@ Usage of ./cmd/mimir/mimir:
     	Maximum number of label names allowed to be queried in a single /api/v1/cardinality/label_values API call. (default 100)
   -querier.lookback-delta duration
     	Time since the last sample after which a time series is considered stale and ignored by expression evaluations. This config option should be set on query-frontend too when query sharding is enabled. (default 5m0s)
+  -querier.max-blocks-per-store-request int
+    	[experimental] Maximum number of blocks that a querier will reference in a single request to a store-gateway. When a request would exceed this, it is split into multiple requests to the same store-gateway. 0 disables the limit.
   -querier.max-concurrent int
     	The number of workers running in each querier process. This setting limits the maximum number of concurrent queries in each querier. The minimum value is four; lower values are ignored and set to the minimum (default 8)
   -querier.max-concurrent-remote-read-queries int

@@ -232,6 +232,7 @@ The following features are currently experimental:
   - Max concurrency for tenant federated queries (`-tenant-federation.max-concurrent`)
   - [Mimir query engine](https://grafana.com/docs/mimir/<MIMIR_VERSION>/references/architecture/mimir-query-engine) (`-querier.query-engine` and `-querier.enable-query-engine-fallback`, and all flags beginning with `-querier.mimir-query-engine`)
   - Maximum estimated memory consumption per query limit (`-querier.max-estimated-memory-consumption-per-query`)
+  - Maximum number of blocks per store-gateway request (`-querier.max-blocks-per-store-request`)
   - Enable the experimental Prometheus feature for delayed name removal (`-querier.enable-delayed-name-removal`)
   - Ignore deletion marks while querying delay (`-blocks-storage.bucket-store.ignore-deletion-marks-while-querying-delay`)
   - Querier ring (all flags beginning with `-querier.ring`)

@@ -4743,6 +4743,12 @@ The `limits` block configures default and per-tenant limits imposed by component
 # CLI flag: -querier.max-estimated-memory-consumption-per-query
 [max_estimated_memory_consumption_per_query: <int> | default = 0]
 
+# (experimental) Maximum number of blocks that a querier will reference in a
+# single request to a store-gateway. When a request would exceed this, it is
+# split into multiple requests to the same store-gateway. 0 disables the limit.
+# CLI flag: -querier.max-blocks-per-store-request
+[max_blocks_per_store_request: <int> | default = 0]
+
 # Limit how long back data (series and metadata) can be queried, up until
 # <lookback> duration ago. This limit is enforced in the query-frontend, querier
 # and ruler for instant, range and remote read queries. For metadata queries

@@ -424,6 +424,7 @@
   "querier.max-fetched-series-per-query": 0,
   "querier.max-fetched-chunk-bytes-per-query": 0,
   "querier.max-estimated-memory-consumption-per-query": 0,
+  "querier.max-blocks-per-store-request": 0,
   "querier.max-query-lookback": 0,
   "querier.max-partial-query-length": 0,
   "querier.max-query-parallelism": 14,