Extend parallel_distributed_insert_select to run INSERT into local table

Before this patch there was:

- parallel_distributed_insert_select=1, that executes:

    INSERT INTO dist_out SELECT ... FROM underlying_dist_in

After this patch there will be:

- parallel_distributed_insert_select=2, that executes:

    INSERT INTO underlying_dist_out SELECT ... FROM underlying_dist_in

And cover the behaviour w/o integration test, by using the following
techincs:
- SYSTEM STOP DISTRIBUTED SENDS
- prefer_localhost_replica=0

docs/en/operations/settings/settings.md

@@ -1571,15 +1571,15 @@ Default value: 16.
 
 ## parallel_distributed_insert_select {#parallel_distributed_insert_select}
 
-Enables parallel distributed `INSERT ... SELECT` query. 
+Enables parallel distributed `INSERT ... SELECT` query.
 
 If we execute `INSERT INTO distributed_table_a SELECT ... FROM distributed_table_b` queries and both tables use the same cluster, and both tables are either [replicated](../../engines/table-engines/mergetree-family/replication.md) or non-replicated, then this query is processed locally on every shard.
 
-
 Possible values:
 
 -   0 — Disabled.
--   1 — Enabled.
+-   1 — `SELECT` will be executed on each shard from underlying table of the distributed engine.
+-   2 — `SELECT` and `INSERT` will be executed on each shard from/to underlying table of the distributed engine.
 
 Default value: 0.
 

src/Core/Settings.h

@@ -108,7 +108,7 @@ class IColumn;
     M(Bool, skip_unavailable_shards, false, "If 1, ClickHouse silently skips unavailable shards and nodes unresolvable through DNS. Shard is marked as unavailable when none of the replicas can be reached.", 0) \
     \
     M(Bool, distributed_group_by_no_merge, false, "Do not merge aggregation states from different servers for distributed query processing - in case it is for certain that there are different keys on different shards.", 0) \
-    M(Bool, parallel_distributed_insert_select, false, "If true, distributed insert select query in the same cluster will be processed on local tables on every shard", 0) \
+    M(UInt64, parallel_distributed_insert_select, 0, "Process distributed INSERT SELECT query in the same cluster on local tables on every shard, if 1 SELECT is executed on each shard, if 2 SELECT and INSERT is executed on each shard", 0) \
     M(Bool, optimize_distributed_group_by_sharding_key, false, "Optimize GROUP BY sharding_key queries (by avodiing costly aggregation on the initiator server).", 0) \
     M(Bool, optimize_skip_unused_shards, false, "Assumes that data is distributed by sharding_key. Optimization to skip unused shards if SELECT query filters by sharding_key.", 0) \
     M(UInt64, force_optimize_skip_unused_shards, 0, "Throw an exception if unused shards cannot be skipped (1 - throw only if the table has the sharding key, 2 - always throw.", 0) \

src/Interpreters/InterpreterInsertQuery.cpp

@@ -29,6 +29,11 @@
 #include <Processors/Transforms/ConvertingTransform.h>
 #include <Processors/Sources/SinkToOutputStream.h>
 
+namespace
+{
+const UInt64 PARALLEL_DISTRIBUTED_INSERT_SELECT_ALL = 2;
+}
+
 
 namespace DB
 {
@@ -157,6 +162,11 @@ BlockIO InterpreterInsertQuery::execute()
         {
             is_distributed_insert_select = true;
 
+            if (settings.parallel_distributed_insert_select == PARALLEL_DISTRIBUTED_INSERT_SELECT_ALL)
+            {
+                new_query->table_id = StorageID(storage_dst->getRemoteDatabaseName(), storage_dst->getRemoteTableName());
+            }
+
             const auto & cluster = storage_src->getCluster();
             const auto & shards_info = cluster->getShardsInfo();
 

tests/queries/0_stateless/01099_parallel_distributed_insert_select.reference

@@ -1,3 +1,4 @@
+parallel_distributed_insert_select=1
 test_shard_localhost
 0
 1
@@ -7,6 +8,30 @@ test_cluster_two_shards_localhost
 1	2
 2	2
 test_cluster_two_shards
+distributed
+local
+0	2
+1	2
+2	2
+distributed
+0	4
+1	4
+2	4
+parallel_distributed_insert_select=2
+test_shard_localhost
+0
+1
+2
+test_cluster_two_shards_localhost
+0	2
+1	2
+2	2
+test_cluster_two_shards
+distributed
+0	4
+1	4
+2	4
+local
 0	2
 1	2
 2	2

tests/queries/0_stateless/01099_parallel_distributed_insert_select.sql

@@ -4,6 +4,7 @@ DROP TABLE IF EXISTS distributed_01099_a;
 DROP TABLE IF EXISTS distributed_01099_b;
 
 SET parallel_distributed_insert_select=1;
+SELECT 'parallel_distributed_insert_select=1';
 
 --
 -- test_shard_localhost
@@ -58,10 +59,98 @@ CREATE TABLE local_01099_b (number UInt64) ENGINE = Log;
 CREATE TABLE distributed_01099_a AS local_01099_a ENGINE = Distributed('test_cluster_two_shards', currentDatabase(), local_01099_a, rand());
 CREATE TABLE distributed_01099_b AS local_01099_b ENGINE = Distributed('test_cluster_two_shards', currentDatabase(), local_01099_b, rand());
 
+SYSTEM STOP DISTRIBUTED SENDS distributed_01099_b;
+SET prefer_localhost_replica=0; -- to require distributed send for local replica too
 INSERT INTO local_01099_a SELECT number from system.numbers limit 3;
 INSERT INTO distributed_01099_b SELECT * from distributed_01099_a;
+SET prefer_localhost_replica=1;
+
+-- distributed sends disabled, 0 rows (since parallel_distributed_insert_select=1)
+SELECT 'distributed';
+SELECT number, count(number) FROM distributed_01099_b group by number order by number;
 SYSTEM FLUSH DISTRIBUTED distributed_01099_b;
 
+SELECT 'local';
+SELECT number, count(number) FROM local_01099_b group by number order by number;
+SELECT 'distributed';
+SELECT number, count(number) FROM distributed_01099_b group by number order by number;
+
+DROP TABLE local_01099_a;
+DROP TABLE local_01099_b;
+DROP TABLE distributed_01099_a;
+DROP TABLE distributed_01099_b;
+
+
+SET parallel_distributed_insert_select=2;
+SELECT 'parallel_distributed_insert_select=2';
+
+--
+-- test_shard_localhost
+--
+
+SELECT 'test_shard_localhost';
+
+CREATE TABLE local_01099_a (number UInt64) ENGINE = Log;
+CREATE TABLE local_01099_b (number UInt64) ENGINE = Log;
+CREATE TABLE distributed_01099_a AS local_01099_a ENGINE = Distributed('test_shard_localhost', currentDatabase(), local_01099_a, rand());
+CREATE TABLE distributed_01099_b AS local_01099_b ENGINE = Distributed('test_shard_localhost', currentDatabase(), local_01099_b, rand());
+
+INSERT INTO local_01099_a SELECT number from system.numbers limit 3;
+INSERT INTO distributed_01099_b SELECT * from distributed_01099_a;
+
+SELECT * FROM distributed_01099_b;
+
+DROP TABLE local_01099_a;
+DROP TABLE local_01099_b;
+DROP TABLE distributed_01099_a;
+DROP TABLE distributed_01099_b;
+
+--
+-- test_cluster_two_shards_localhost
+--
+
+SELECT 'test_cluster_two_shards_localhost';
+
+-- Log engine will lead to deadlock:
+--     DB::Exception: std::system_error: Resource deadlock avoided.
+-- So use MergeTree instead.
+CREATE TABLE local_01099_a (number UInt64) ENGINE = MergeTree() ORDER BY number;
+CREATE TABLE local_01099_b (number UInt64) ENGINE = MergeTree() ORDER BY number;
+CREATE TABLE distributed_01099_a AS local_01099_a ENGINE = Distributed('test_cluster_two_shards_localhost', currentDatabase(), local_01099_a, rand());
+CREATE TABLE distributed_01099_b AS local_01099_b ENGINE = Distributed('test_cluster_two_shards_localhost', currentDatabase(), local_01099_b, rand());
+
+INSERT INTO local_01099_a SELECT number from system.numbers limit 3;
+INSERT INTO distributed_01099_b SELECT * from distributed_01099_a;
+
+SELECT number, count(number) FROM local_01099_b group by number order by number;
+
+DROP TABLE local_01099_a;
+DROP TABLE local_01099_b;
+DROP TABLE distributed_01099_a;
+DROP TABLE distributed_01099_b;
+
+--
+-- test_cluster_two_shards
+--
+
+SELECT 'test_cluster_two_shards';
+
+CREATE TABLE local_01099_a (number UInt64) ENGINE = MergeTree() ORDER BY number;
+CREATE TABLE local_01099_b (number UInt64) ENGINE = MergeTree() ORDER BY number;
+CREATE TABLE distributed_01099_a AS local_01099_a ENGINE = Distributed('test_cluster_two_shards', currentDatabase(), local_01099_a, rand());
+CREATE TABLE distributed_01099_b AS local_01099_b ENGINE = Distributed('test_cluster_two_shards', currentDatabase(), local_01099_b, rand());
+
+SYSTEM STOP DISTRIBUTED SENDS distributed_01099_b;
+SET prefer_localhost_replica=0; -- to require distributed send for local replica too
+INSERT INTO local_01099_a SELECT number from system.numbers limit 3;
+INSERT INTO distributed_01099_b SELECT * from distributed_01099_a;
+SET prefer_localhost_replica=1;
+
+-- distributed sends disabled, but they are not required, since insert is done into local table.
+-- (since parallel_distributed_insert_select=2)
+SELECT 'distributed';
+SELECT number, count(number) FROM distributed_01099_b group by number order by number;
+SELECT 'local';
 SELECT number, count(number) FROM local_01099_b group by number order by number;
 
 DROP TABLE local_01099_a;