Better thread pool

src/Core/Settings.h

@@ -155,8 +155,6 @@ static constexpr UInt64 operator""_GiB(unsigned long long value)
     \
     M(Bool, allow_experimental_parallel_reading_from_replicas, false, "If true, ClickHouse will send a SELECT query to all replicas of a table. It will work for any kind on MergeTree table.", 0) \
     \
-    M(UInt64, system_replicas_fetch_threads, 16, "The maximum number of threads to fetch data for system.replicas table.", 0) \
-    \
     M(Bool, skip_unavailable_shards, false, "If true, 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(UInt64, parallel_distributed_insert_select, 0, "Process distributed INSERT SELECT query in the same cluster on local tables on every shard; if set to 1 - SELECT is executed on each shard; if set to 2 - SELECT and INSERT are executed on each shard", 0) \

src/Storages/System/StorageSystemReplicas.cpp

@@ -10,6 +10,7 @@
 #include <Common/typeid_cast.h>
 #include <Databases/IDatabase.h>
 #include <Processors/Sources/SourceFromSingleChunk.h>
+#include <Common/getNumberOfPhysicalCPUCores.h>
 
 
 namespace DB
@@ -151,30 +152,46 @@ Pipe StorageSystemReplicas::read(
 
     MutableColumns res_columns = storage_snapshot->metadata->getSampleBlock().cloneEmptyColumns();
 
-    auto settings = context->getSettingsRef();
-    size_t thread_pool_size = settings.system_replicas_fetch_threads;
+    size_t tables_size = col_database->size();
+    size_t thread_pool_size = std::min(tables_size, static_cast<size_t>(getNumberOfPhysicalCPUCores()));
 
+    auto settings = context->getSettingsRef();
     if (settings.max_threads != 0)
         thread_pool_size = std::min(thread_pool_size, static_cast<size_t>(settings.max_threads));
 
     ThreadPool thread_pool(thread_pool_size);
+    std::atomic<bool> error_flag = false;
+    Exception exception;
 
-    size_t tables_size = col_database->size();
     std::vector<StorageReplicatedMergeTree::Status> statuses(tables_size);
 
     for (size_t i = 0; i < tables_size; ++i)
     {
-        thread_pool.scheduleOrThrowOnError([i, &statuses, &replicated_tables, &col_database, &col_table, &with_zk_fields]
+        thread_pool.scheduleOrThrowOnError([&, i=i]
         {
-            dynamic_cast<StorageReplicatedMergeTree &>(
+            try
+            {
+                dynamic_cast<StorageReplicatedMergeTree &>(
                 *replicated_tables
                     [(*col_database)[i].safeGet<const String &>()]
                     [(*col_table)[i].safeGet<const String &>()]).getStatus(statuses[i], with_zk_fields);
+            }
+            catch (...)
+            {
+                tryLogCurrentException("system.replicas", "Failed to fetch system.replicas data");
+
+                /// We capture one of the exceptions to be thrown later
+                if (!error_flag.exchange(true))
+                    exception = Exception(getCurrentExceptionCode(), getCurrentExceptionMessage(false));
+            }
         });
     }
 
     thread_pool.wait();
 
+    if (error_flag)
+        throw exception;
+
     for (const auto & status: statuses)
     {
         size_t col_num = 3;