Merge pull request #3851 from abyss7/issue-3592

Refactor constant folding

.gitignore

@@ -248,3 +248,6 @@ website/package-lock.json
 
 # Ignore files for locally disabled tests
 /dbms/tests/queries/**/*.disabled
+
+# cquery cache
+/.cquery-cache

dbms/programs/server/config.xml

@@ -187,6 +187,20 @@
                 </replica>
             </shard>
         </test_shard_localhost_secure>
+        <test_unavailable_shard>
+            <shard>
+                <replica>
+                    <host>localhost</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+            <shard>
+                <replica>
+                    <host>localhost</host>
+                    <port>1</port>
+                </replica>
+            </shard>
+        </test_unavailable_shard>
     </remote_servers>
 
 

dbms/src/Core/Block.h

@@ -97,8 +97,8 @@ public:
     /// Approximate number of allocated bytes in memory - for profiling and limits.
     size_t allocatedBytes() const;
 
-    operator bool() const { return !data.empty(); }
-    bool operator!() const { return data.empty(); }
+    operator bool() const { return !!columns(); }
+    bool operator!() const { return !this->operator bool(); }
 
     /** Get a list of column names separated by commas. */
     std::string dumpNames() const;

dbms/src/Interpreters/Cluster.cpp

@@ -397,14 +397,24 @@ void Cluster::initMisc()
 
 std::unique_ptr<Cluster> Cluster::getClusterWithSingleShard(size_t index) const
 {
-    return std::unique_ptr<Cluster>{ new Cluster(*this, index) };
+    return std::unique_ptr<Cluster>{ new Cluster(*this, {index}) };
 }
 
-Cluster::Cluster(const Cluster & from, size_t index)
-    : shards_info{from.shards_info[index]}
+std::unique_ptr<Cluster> Cluster::getClusterWithMultipleShards(const std::vector<size_t> & indices) const
 {
-    if (!from.addresses_with_failover.empty())
-        addresses_with_failover.emplace_back(from.addresses_with_failover[index]);
+    return std::unique_ptr<Cluster>{ new Cluster(*this, indices) };
+}
+
+Cluster::Cluster(const Cluster & from, const std::vector<size_t> & indices)
+    : shards_info{}
+{
+    for (size_t index : indices)
+    {
+        shards_info.emplace_back(from.shards_info.at(index));
+
+        if (!from.addresses_with_failover.empty())
+            addresses_with_failover.emplace_back(from.addresses_with_failover.at(index));
+    }
 
     initMisc();
 }

dbms/src/Interpreters/Cluster.h

@@ -143,6 +143,9 @@ public:
     /// Get a subcluster consisting of one shard - index by count (from 0) of the shard of this cluster.
     std::unique_ptr<Cluster> getClusterWithSingleShard(size_t index) const;
 
+    /// Get a subcluster consisting of one or multiple shards - indexes by count (from 0) of the shard of this cluster.
+    std::unique_ptr<Cluster> getClusterWithMultipleShards(const std::vector<size_t> & indices) const;
+
 private:
     using SlotToShard = std::vector<UInt64>;
     SlotToShard slot_to_shard;
@@ -153,8 +156,8 @@ public:
 private:
     void initMisc();
 
-    /// For getClusterWithSingleShard implementation.
-    Cluster(const Cluster & from, size_t index);
+    /// For getClusterWithMultipleShards implementation.
+    Cluster(const Cluster & from, const std::vector<size_t> & indices);
 
     String hash_of_addresses;
     /// Description of the cluster shards.

dbms/src/Interpreters/Settings.h

@@ -89,6 +89,7 @@ struct Settings
     M(SettingBool, skip_unavailable_shards, false, "Silently skip unavailable shards.") \
     \
     M(SettingBool, 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.") \
+    M(SettingBool, distributed_optimize_skip_select_on_unused_shards, false, "Assumes that data is distributed by sharding_key. Optimization to skip unused shards if SELECT query filters by sharding_key.") \
     \
     M(SettingUInt64, merge_tree_min_rows_for_concurrent_read, (20 * 8192), "If at least as many lines are read from one file, the reading can be parallelized.") \
     M(SettingUInt64, merge_tree_min_rows_for_seek, 0, "You can skip reading more than that number of rows at the price of one seek per file.") \

dbms/src/Interpreters/evaluateConstantExpression.cpp

@@ -1,18 +1,20 @@
-#include <Core/Block.h>
+#include <Interpreters/evaluateConstantExpression.h>
+
 #include <Columns/ColumnConst.h>
 #include <Columns/ColumnsNumber.h>
-#include <Parsers/ASTIdentifier.h>
-#include <Parsers/ASTLiteral.h>
-#include <Parsers/ASTFunction.h>
-#include <Parsers/ExpressionElementParsers.h>
+#include <Core/Block.h>
 #include <DataTypes/DataTypesNumber.h>
 #include <Interpreters/Context.h>
-#include <Interpreters/SyntaxAnalyzer.h>
-#include <Interpreters/ExpressionAnalyzer.h>
+#include <Interpreters/convertFieldToType.h>
 #include <Interpreters/ExpressionActions.h>
-#include <Interpreters/evaluateConstantExpression.h>
-#include <Common/typeid_cast.h>
+#include <Interpreters/ExpressionAnalyzer.h>
+#include <Interpreters/SyntaxAnalyzer.h>
+#include <Parsers/ASTFunction.h>
+#include <Parsers/ASTIdentifier.h>
+#include <Parsers/ASTLiteral.h>
+#include <Parsers/ExpressionElementParsers.h>
 #include <TableFunctions/TableFunctionFactory.h>
+#include <Common/typeid_cast.h>
 
 
 namespace DB
@@ -77,4 +79,236 @@ ASTPtr evaluateConstantExpressionOrIdentifierAsLiteral(const ASTPtr & node, cons
     return evaluateConstantExpressionAsLiteral(node, context);
 }
 
+namespace
+{
+    using Conjunction = ColumnsWithTypeAndName;
+    using Disjunction = std::vector<Conjunction>;
+
+    Disjunction analyzeEquals(const ASTIdentifier * identifier, const ASTLiteral * literal, const ExpressionActionsPtr & expr)
+    {
+        if (!identifier || !literal)
+        {
+            return {};
+        }
+
+        for (const auto & name_and_type : expr->getRequiredColumnsWithTypes())
+        {
+            const auto & name = name_and_type.name;
+            const auto & type = name_and_type.type;
+
+            if (name == identifier->name)
+            {
+                ColumnWithTypeAndName column;
+                // FIXME: what to do if field is not convertable?
+                column.column = type->createColumnConst(1, convertFieldToType(literal->value, *type));
+                column.name = name;
+                column.type = type;
+                return {{std::move(column)}};
+            }
+        }
+
+        return {};
+    }
+
+    Disjunction andDNF(const Disjunction & left, const Disjunction & right)
+    {
+        if (left.empty())
+        {
+            return right;
+        }
+
+        Disjunction result;
+
+        for (const auto & conjunct1 : left)
+        {
+            for (const auto & conjunct2 : right)
+            {
+                Conjunction new_conjunct{conjunct1};
+                new_conjunct.insert(new_conjunct.end(), conjunct2.begin(), conjunct2.end());
+                result.emplace_back(new_conjunct);
+            }
+        }
+
+        return result;
+    }
+
+    Disjunction analyzeFunction(const ASTFunction * fn, const ExpressionActionsPtr & expr)
+    {
+        if (!fn)
+        {
+            return {};
+        }
+
+        // TODO: enumerate all possible function names!
+
+        if (fn->name == "equals")
+        {
+            const auto * left = fn->arguments->children.front().get();
+            const auto * right = fn->arguments->children.back().get();
+            const auto * identifier = typeid_cast<const ASTIdentifier *>(left) ? typeid_cast<const ASTIdentifier *>(left)
+                                                                               : typeid_cast<const ASTIdentifier *>(right);
+            const auto * literal = typeid_cast<const ASTLiteral *>(left) ? typeid_cast<const ASTLiteral *>(left)
+                                                                         : typeid_cast<const ASTLiteral *>(right);
+
+            return analyzeEquals(identifier, literal, expr);
+        }
+        else if (fn->name == "in")
+        {
+            const auto * left = fn->arguments->children.front().get();
+            const auto * right = fn->arguments->children.back().get();
+            const auto * identifier = typeid_cast<const ASTIdentifier *>(left);
+            const auto * inner_fn = typeid_cast<const ASTFunction *>(right);
+
+            if (!inner_fn)
+            {
+                return {};
+            }
+
+            const auto * tuple = typeid_cast<const ASTExpressionList *>(inner_fn->children.front().get());
+
+            if (!tuple)
+            {
+                return {};
+            }
+
+            Disjunction result;
+
+            for (const auto & child : tuple->children)
+            {
+                const auto * literal = typeid_cast<const ASTLiteral *>(child.get());
+                const auto dnf = analyzeEquals(identifier, literal, expr);
+
+                if (dnf.empty())
+                {
+                    return {};
+                }
+
+                result.insert(result.end(), dnf.begin(), dnf.end());
+            }
+
+            return result;
+        }
+        else if (fn->name == "or")
+        {
+            const auto * args = typeid_cast<const ASTExpressionList *>(fn->children.front().get());
+
+            if (!args)
+            {
+                return {};
+            }
+
+            Disjunction result;
+
+            for (const auto & arg : args->children)
+            {
+                const auto dnf = analyzeFunction(typeid_cast<const ASTFunction *>(arg.get()), expr);
+
+                if (dnf.empty())
+                {
+                    return {};
+                }
+
+                result.insert(result.end(), dnf.begin(), dnf.end());
+            }
+
+            return result;
+        }
+        else if (fn->name == "and")
+        {
+            const auto * args = typeid_cast<const ASTExpressionList *>(fn->children.front().get());
+
+            if (!args)
+            {
+                return {};
+            }
+
+            Disjunction result;
+
+            for (const auto & arg : args->children)
+            {
+                const auto dnf = analyzeFunction(typeid_cast<const ASTFunction *>(arg.get()), expr);
+
+                if (dnf.empty())
+                {
+                    continue;
+                }
+
+                result = andDNF(result, dnf);
+            }
+
+            return result;
+        }
+
+        return {};
+    }
+}
+
+std::optional<Blocks> evaluateExpressionOverConstantCondition(const ASTPtr & node, const ExpressionActionsPtr & target_expr)
+{
+    Blocks result;
+
+    // TODO: `node` may be always-false literal.
+
+    if (const auto fn = typeid_cast<const ASTFunction *>(node.get()))
+    {
+        const auto dnf = analyzeFunction(fn, target_expr);
+
+        if (dnf.empty())
+        {
+            return {};
+        }
+
+        auto hasRequiredColumns = [&target_expr](const Block & block) -> bool
+        {
+            for (const auto & name : target_expr->getRequiredColumns())
+            {
+                bool hasColumn = false;
+                for (const auto & column_name : block.getNames())
+                {
+                    if (column_name == name)
+                    {
+                        hasColumn = true;
+                        break;
+                    }
+                }
+
+                if (!hasColumn)
+                    return false;
+            }
+
+            return true;
+        };
+
+        for (const auto & conjunct : dnf)
+        {
+            Block block(conjunct);
+
+            // Block should contain all required columns from `target_expr`
+            if (!hasRequiredColumns(block))
+            {
+                return {};
+            }
+
+            target_expr->execute(block);
+
+            if (block.rows() == 1)
+            {
+                result.push_back(block);
+            }
+            else if (block.rows() == 0)
+            {
+                // filter out cases like "WHERE a = 1 AND a = 2"
+                continue;
+            }
+            else
+            {
+                // FIXME: shouldn't happen
+                return {};
+            }
+        }
+    }
+
+    return {result};
+}
+
 }

dbms/src/Interpreters/evaluateConstantExpression.h

@@ -1,17 +1,22 @@
 #pragma once
 
-#include <memory>
+#include <Core/Block.h>
 #include <Core/Field.h>
 #include <Parsers/IAST.h>
 #include <Parsers/IParser.h>
 
+#include <memory>
+#include <optional>
+
 
 namespace DB
 {
 
 class Context;
+class ExpressionActions;
 class IDataType;
 
+using ExpressionActionsPtr = std::shared_ptr<ExpressionActions>;
 
 /** Evaluate constant expression and its type.
   * Used in rare cases - for elements of set for IN, for data to INSERT.
@@ -20,17 +25,24 @@ class IDataType;
 std::pair<Field, std::shared_ptr<const IDataType>> evaluateConstantExpression(const ASTPtr & node, const Context & context);
 
 
-/** Evaluate constant expression
-  *  and returns ASTLiteral with its value.
+/** Evaluate constant expression and returns ASTLiteral with its value.
   */
 ASTPtr evaluateConstantExpressionAsLiteral(const ASTPtr & node, const Context & context);
 
 
-/** Evaluate constant expression
-  *  and returns ASTLiteral with its value.
+/** Evaluate constant expression and returns ASTLiteral with its value.
   * Also, if AST is identifier, then return string literal with its name.
   * Useful in places where some name may be specified as identifier, or as result of a constant expression.
   */
 ASTPtr evaluateConstantExpressionOrIdentifierAsLiteral(const ASTPtr & node, const Context & context);
 
+/** Try to fold condition to countable set of constant values.
+  * @param condition a condition that we try to fold.
+  * @param target_expr expression evaluated over a set of constants.
+  * @return optional blocks each with a single row and a single column for target expression,
+  *         or empty blocks if condition is always false,
+  *         or nothing if condition can't be folded to a set of constants.
+  */
+std::optional<Blocks> evaluateExpressionOverConstantCondition(const ASTPtr & condition, const ExpressionActionsPtr & target_expr);
+
 }

dbms/src/Storages/MergeTree/KeyCondition.cpp

@@ -313,7 +313,7 @@ bool KeyCondition::addCondition(const String & column, const Range & range)
     return true;
 }
 
-/** Computes value of constant expression and it data type.
+/** Computes value of constant expression and its data type.
   * Returns false, if expression isn't constant.
   */
 static bool getConstant(const ASTPtr & expr, Block & block_with_constants, Field & out_value, DataTypePtr & out_type)

dbms/src/Storages/MergeTree/KeyCondition.h

@@ -253,7 +253,7 @@ public:
     /// Get the maximum number of the key element used in the condition.
     size_t getMaxKeyColumn() const;
 
-    /// Impose an additional condition: the value in the column column must be in the `range` range.
+    /// Impose an additional condition: the value in the column `column` must be in the range `range`.
     /// Returns whether there is such a column in the key.
     bool addCondition(const String & column, const Range & range);
 

dbms/src/Storages/StorageDistributed.cpp

@@ -1,38 +1,41 @@
+#include <Storages/StorageDistributed.h>
+
 #include <DataStreams/OneBlockInputStream.h>
 #include <DataStreams/materializeBlock.h>
 
 #include <Databases/IDatabase.h>
 
 #include <DataTypes/DataTypeFactory.h>
+#include <DataTypes/DataTypesNumber.h>
 
-#include <Storages/StorageDistributed.h>
-#include <Storages/Distributed/DistributedBlockOutputStream.h>
 #include <Storages/Distributed/DirectoryMonitor.h>
+#include <Storages/Distributed/DistributedBlockOutputStream.h>
 #include <Storages/StorageFactory.h>
 
 #include <Common/Macros.h>
 #include <Common/escapeForFileName.h>
 #include <Common/typeid_cast.h>
 
-#include <Parsers/ASTInsertQuery.h>
-#include <Parsers/ASTSelectQuery.h>
-#include <Parsers/TablePropertiesQueriesASTs.h>
-#include <Parsers/ParserAlterQuery.h>
-#include <Parsers/parseQuery.h>
-#include <Parsers/ASTLiteral.h>
-#include <Parsers/ASTExpressionList.h>
-#include <Parsers/ASTTablesInSelectQuery.h>
 #include <Parsers/ASTDropQuery.h>
+#include <Parsers/ASTExpressionList.h>
 #include <Parsers/ASTIdentifier.h>
+#include <Parsers/ASTInsertQuery.h>
+#include <Parsers/ASTLiteral.h>
+#include <Parsers/ASTSelectQuery.h>
+#include <Parsers/ASTTablesInSelectQuery.h>
+#include <Parsers/ParserAlterQuery.h>
+#include <Parsers/TablePropertiesQueriesASTs.h>
+#include <Parsers/parseQuery.h>
 
-#include <Interpreters/InterpreterSelectQuery.h>
+#include <Interpreters/ClusterProxy/SelectStreamFactory.h>
+#include <Interpreters/ClusterProxy/executeQuery.h>
+#include <Interpreters/ExpressionAnalyzer.h>
 #include <Interpreters/InterpreterAlterQuery.h>
 #include <Interpreters/InterpreterDescribeQuery.h>
+#include <Interpreters/InterpreterSelectQuery.h>
 #include <Interpreters/SyntaxAnalyzer.h>
-#include <Interpreters/ExpressionAnalyzer.h>
+#include <Interpreters/createBlockSelector.h>
 #include <Interpreters/evaluateConstantExpression.h>
-#include <Interpreters/ClusterProxy/executeQuery.h>
-#include <Interpreters/ClusterProxy/SelectStreamFactory.h>
 #include <Interpreters/getClusterName.h>
 
 #include <Core/Field.h>
@@ -58,6 +61,7 @@ namespace ErrorCodes
     extern const int INFINITE_LOOP;
     extern const int TYPE_MISMATCH;
     extern const int NO_SUCH_COLUMN_IN_TABLE;
+    extern const int TOO_MANY_ROWS;
 }
 
 
@@ -133,6 +137,29 @@ void initializeFileNamesIncrement(const std::string & path, SimpleIncrement & in
         increment.set(getMaximumFileNumber(path));
 }
 
+/// the same as DistributedBlockOutputStream::createSelector, should it be static?
+IColumn::Selector createSelector(const ClusterPtr cluster, const ColumnWithTypeAndName & result)
+{
+    const auto & slot_to_shard = cluster->getSlotToShard();
+
+#define CREATE_FOR_TYPE(TYPE)                                   \
+    if (typeid_cast<const DataType##TYPE *>(result.type.get())) \
+        return createBlockSelector<TYPE>(*result.column, slot_to_shard);
+
+    CREATE_FOR_TYPE(UInt8)
+    CREATE_FOR_TYPE(UInt16)
+    CREATE_FOR_TYPE(UInt32)
+    CREATE_FOR_TYPE(UInt64)
+    CREATE_FOR_TYPE(Int8)
+    CREATE_FOR_TYPE(Int16)
+    CREATE_FOR_TYPE(Int32)
+    CREATE_FOR_TYPE(Int64)
+
+#undef CREATE_FOR_TYPE
+
+    throw Exception{"Sharding key expression does not evaluate to an integer type", ErrorCodes::TYPE_MISMATCH};
+}
+
 }
 
 
@@ -267,6 +294,14 @@ BlockInputStreams StorageDistributed::read(
         : ClusterProxy::SelectStreamFactory(
             header, processed_stage, QualifiedTableName{remote_database, remote_table}, context.getExternalTables());
 
+    if (settings.distributed_optimize_skip_select_on_unused_shards)
+    {
+        auto smaller_cluster = skipUnusedShards(cluster, query_info);
+
+        if (smaller_cluster)
+            cluster = smaller_cluster;
+    }
+
     return ClusterProxy::executeQuery(
         select_stream_factory, cluster, modified_query_ast, context, settings);
 }
@@ -425,6 +460,41 @@ void StorageDistributed::ClusterNodeData::shutdownAndDropAllData()
     directory_monitor->shutdownAndDropAllData();
 }
 
+/// Returns a new cluster with fewer shards if constant folding for `sharding_key_expr` is possible
+/// using constraints from "WHERE" condition, otherwise returns `nullptr`
+ClusterPtr StorageDistributed::skipUnusedShards(ClusterPtr cluster, const SelectQueryInfo & query_info)
+{
+    const auto & select = typeid_cast<ASTSelectQuery &>(*query_info.query);
+
+    if (!select.where_expression)
+    {
+        return nullptr;
+    }
+
+    const auto & blocks = evaluateExpressionOverConstantCondition(select.where_expression, sharding_key_expr);
+
+    // Can't get definite answer if we can skip any shards
+    if (!blocks)
+    {
+        return nullptr;
+    }
+
+    std::set<int> shards;
+
+    for (const auto & block : *blocks)
+    {
+        if (!block.has(sharding_key_column_name))
+            throw Exception("sharding_key_expr should evaluate as a single row", ErrorCodes::TOO_MANY_ROWS);
+
+        const auto result = block.getByName(sharding_key_column_name);
+        const auto selector = createSelector(cluster, result);
+
+        shards.insert(selector.begin(), selector.end());
+    }
+
+    return cluster->getClusterWithMultipleShards({shards.begin(), shards.end()});
+}
+
 
 void registerStorageDistributed(StorageFactory & factory)
 {

dbms/src/Storages/StorageDistributed.h

@@ -166,6 +166,8 @@ protected:
         const ASTPtr & sharding_key_,
         const String & data_path_,
         bool attach);
+
+    ClusterPtr skipUnusedShards(ClusterPtr cluster, const SelectQueryInfo & query_info);
 };
 
 }

dbms/tests/queries/0_stateless/00754_distributed_optimize_skip_select_on_unused_shards.reference

@@ -0,0 +1,16 @@
+OK
+OK
+1
+OK
+0
+4
+2
+1
+1
+1
+4
+OK
+OK
+OK
+OK
+OK

dbms/tests/queries/0_stateless/00754_distributed_optimize_skip_select_on_unused_shards.sh

@@ -0,0 +1,105 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. $CURDIR/../shell_config.sh
+
+${CLICKHOUSE_CLIENT} --query "DROP TABLE IF EXISTS test.mergetree;"
+${CLICKHOUSE_CLIENT} --query "DROP TABLE IF EXISTS test.distributed;"
+
+${CLICKHOUSE_CLIENT} --query "CREATE TABLE test.mergetree (a Int64, b Int64, c Int64) ENGINE = MergeTree ORDER BY (a, b);"
+${CLICKHOUSE_CLIENT} --query "CREATE TABLE test.distributed AS test.mergetree ENGINE = Distributed(test_unavailable_shard, test, mergetree, jumpConsistentHash(a+b, 2));"
+
+${CLICKHOUSE_CLIENT} --query "INSERT INTO test.mergetree VALUES (0, 0, 0);"
+${CLICKHOUSE_CLIENT} --query "INSERT INTO test.mergetree VALUES (1, 0, 0);"
+${CLICKHOUSE_CLIENT} --query "INSERT INTO test.mergetree VALUES (0, 1, 1);"
+${CLICKHOUSE_CLIENT} --query "INSERT INTO test.mergetree VALUES (1, 1, 1);"
+
+# Should fail because second shard is unavailable
+${CLICKHOUSE_CLIENT} --query "SELECT count(*) FROM test.distributed;" 2>&1 \
+| fgrep -q "All connection tries failed" && echo 'OK' || echo 'FAIL'
+
+# Should fail without setting `distributed_optimize_skip_select_on_unused_shards`
+${CLICKHOUSE_CLIENT} --query "SELECT count(*) FROM test.distributed WHERE a = 0 AND b = 0;" 2>&1 \
+| fgrep -q "All connection tries failed" && echo 'OK' || echo 'FAIL'
+
+# Should pass now
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 AND b = 0;
+"
+
+# Should still fail because of matching unavailable shard
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 2 AND b = 2;
+" 2>&1 \ | fgrep -q "All connection tries failed" && echo 'OK' || echo 'FAIL'
+
+# Try more complext expressions for constant folding - all should pass.
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 1 AND a = 0 AND b = 0;
+"
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a IN (0, 1) AND b IN (0, 1);
+"
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 AND b = 0 OR a = 1 AND b = 1;
+"
+
+# TODO: should pass one day.
+#${CLICKHOUSE_CLIENT} -n --query="
+#    SET distributed_optimize_skip_select_on_unused_shards = 1;
+#    SELECT count(*) FROM test.distributed WHERE a = 0 AND b >= 0 AND b <= 1;
+#"
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 AND b = 0 AND c = 0;
+"
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 AND b = 0 AND c != 10;
+"
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 AND b = 0 AND (a+b)*b != 12;
+"
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE (a = 0 OR a = 1) AND (b = 0 OR b = 1);
+"
+
+# These ones should fail.
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 AND b <= 1;
+" 2>&1 \ | fgrep -q "All connection tries failed" && echo 'OK' || echo 'FAIL'
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 AND c = 0;
+" 2>&1 \ | fgrep -q "All connection tries failed" && echo 'OK' || echo 'FAIL'
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 OR a = 1 AND b = 0;
+" 2>&1 \ | fgrep -q "All connection tries failed" && echo 'OK' || echo 'FAIL'
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 AND b = 0 OR a = 2 AND b = 2;
+" 2>&1 \ | fgrep -q "All connection tries failed" && echo 'OK' || echo 'FAIL'
+
+${CLICKHOUSE_CLIENT} -n --query="
+    SET distributed_optimize_skip_select_on_unused_shards = 1;
+    SELECT count(*) FROM test.distributed WHERE a = 0 AND b = 0 OR c = 0;
+" 2>&1 \ | fgrep -q "All connection tries failed" && echo 'OK' || echo 'FAIL'

dbms/tests/server-test.xml

@@ -53,6 +53,20 @@
     <timezone>Europe/Moscow</timezone>
     <remote_servers incl="clickhouse_remote_servers" >
         <!-- Test only shard config for testing distributed storage -->
+        <test_unavailable_shard>
+            <shard>
+                <replica>
+                    <host>localhost</host>
+                    <port>59000</port>
+                </replica>
+            </shard>
+            <shard>
+                <replica>
+                    <host>localhost</host>
+                    <port>1</port>
+                </replica>
+            </shard>
+        </test_unavailable_shard>
         <test_shard_localhost>
             <shard>
                 <replica>