YT-21484: Consider partition sizes in coordination by sample keys

commit_hash:56feb3c28bc4d963431d8f70140fd2ae2406da78

Author: lukyan

yt/yt/client/api/client_common.h

@@ -144,6 +144,8 @@ struct TSelectRowsOptionsBase
     ui64 RangeExpansionLimit = 200000;
     //! Limits maximum parallel subqueries.
     int MaxSubqueries = std::numeric_limits<int>::max();
+    //! Limits parallel subqueries by row count.
+    ui64 MinRowCountPerSubquery = 100'000;
     //! Path in Cypress with UDFs.
     std::optional<TString> UdfRegistryPath;
     //! If |true| then logging is more verbose.

yt/yt/client/driver/command-inl.h

@@ -392,6 +392,14 @@ void TSelectRowsCommandBase<
         .GreaterThan(0)
         .Optional(/*init*/ false);
 
+    registrar.template ParameterWithUniversalAccessor<ui64>(
+        "min_row_count_per_subquery",
+        [] (TThis* command) -> auto& {
+            return command->Options.MinRowCountPerSubquery;
+        })
+        .GreaterThan(0)
+        .Optional(/*init*/ false);
+
     registrar.template ParameterWithUniversalAccessor<std::optional<TString>>(
         "udf_registry_path",
         [] (TThis* command) -> auto& {

yt/yt_proto/yt/client/api/rpc_proxy/proto/api_service.proto

@@ -638,6 +638,7 @@ message TReqSelectRows
     optional int32 execution_backend = 23; // EExecutionBackend
     optional NYT.NTableClient.NProto.TVersionedReadOptions versioned_read_options = 25;
     optional bool use_lookup_cache = 26;
+    optional int64 min_row_count_per_subquery = 27;
 
     optional TSuppressableAccessTrackingOptions suppressable_access_tracking_options = 104;
 
@@ -836,6 +837,7 @@ message TReqExplainQuery
     optional string execution_pool = 11;
     optional bool new_range_inference = 12;
     optional int32 syntax_version = 13 [default = 1];
+    optional int64 min_row_count_per_subquery = 14;
 }
 
 message TRspExplainQuery