Added "os_thread_priority" setting

dbms/programs/server/Server.cpp

@@ -533,10 +533,18 @@ int Server::main(const std::vector<std::string> & /*args*/)
     if (!TaskStatsInfoGetter::checkPermissions())
     {
         LOG_INFO(log, "It looks like the process has no CAP_NET_ADMIN capability, 'taskstats' performance statistics will be disabled."
-                      " It could happen due to incorrect ClickHouse package installation."
-                      " You could resolve the problem manually with 'sudo setcap cap_net_admin=+ep /usr/bin/clickhouse'."
-                      " Note that it will not work on 'nosuid' mounted filesystems."
-                      " It also doesn't work if you run clickhouse-server inside network namespace as it happens in some containers.");
+            " It could happen due to incorrect ClickHouse package installation."
+            " You could resolve the problem manually with 'sudo setcap cap_net_admin=+ep /usr/bin/clickhouse'."
+            " Note that it will not work on 'nosuid' mounted filesystems."
+            " It also doesn't work if you run clickhouse-server inside network namespace as it happens in some containers.");
+    }
+
+    if (!hasLinuxCapability(CAP_SYS_NICE))
+    {
+        LOG_INFO(log, "It looks like the process has no CAP_SYS_NICE capability, the setting 'os_thread_nice' will have no effect."
+            " It could happen due to incorrect ClickHouse package installation."
+            " You could resolve the problem manually with 'sudo setcap cap_sys_nice=+ep /usr/bin/clickhouse'."
+            " Note that it will not work on 'nosuid' mounted filesystems.");
     }
 #else
     LOG_INFO(log, "TaskStats is not implemented for this OS. IO accounting will be disabled.");

dbms/src/Common/ErrorCodes.cpp

@@ -433,6 +433,7 @@ namespace ErrorCodes
     extern const int UNKNOWN_QUERY_PARAMETER = 456;
     extern const int BAD_QUERY_PARAMETER = 457;
     extern const int CANNOT_UNLINK = 458;
+    extern const int CANNOT_SET_THREAD_PRIORITY = 459;
 
     extern const int KEEPER_EXCEPTION = 999;
     extern const int POCO_EXCEPTION = 1000;

dbms/src/Common/ThreadStatus.h

@@ -86,6 +86,8 @@ public:
     UInt32 thread_number = 0;
     /// Linux's PID (or TGID) (the same id is shown by ps util)
     Int32 os_thread_id = -1;
+    /// Also called "nice" value. If it was changed to non-zero (when attaching query) - will be reset to zero when query is detached.
+    Int32 os_thread_priority = 0;
 
     /// TODO: merge them into common entity
     ProfileEvents::Counters performance_counters{VariableContext::Thread};

dbms/src/Core/Settings.h

@@ -132,6 +132,7 @@ struct Settings : public SettingsCollection<Settings>
     M(SettingInt64, network_zstd_compression_level, 1, "Allows you to select the level of ZSTD compression.") \
     \
     M(SettingUInt64, priority, 0, "Priority of the query. 1 - the highest, higher value - lower priority; 0 - do not use priorities.") \
+    M(SettingInt64, os_thread_priority, 0, "If non zero - set corresponding 'nice' value for query processing threads. Can be used to adjust query priority for OS scheduler.") \
     \
     M(SettingBool, log_queries, 0, "Log requests and write the log to the system table.") \
     \

dbms/src/Interpreters/ThreadStatusExt.cpp

@@ -6,6 +6,13 @@
 #include <Interpreters/QueryThreadLog.h>
 #include <Interpreters/ProcessList.h>
 
+#if defined(__linux__)
+#include <sys/time.h>
+#include <sys/resource.h>
+
+#include <Common/hasLinuxCapability.h>
+#endif
+
 
 /// Implement some methods of ThreadStatus and CurrentThread here to avoid extra linking dependencies in clickhouse_common_io
 /// TODO It doesn't make sense.
@@ -13,6 +20,12 @@
 namespace DB
 {
 
+namespace ErrorCodes
+{
+    extern const int CANNOT_SET_THREAD_PRIORITY;
+}
+
+
 void ThreadStatus::attachQueryContext(Context & query_context_)
 {
     query_context = &query_context_;
@@ -93,6 +106,23 @@ void ThreadStatus::attachQuery(const ThreadGroupStatusPtr & thread_group_, bool
         thread_group->thread_numbers.emplace_back(thread_number);
     }
 
+#if defined(__linux__)
+    /// Set "nice" value if required.
+    if (query_context)
+    {
+        Int32 new_os_thread_priority = query_context->getSettingsRef().os_thread_priority;
+        if (new_os_thread_priority && hasLinuxCapability(CAP_SYS_NICE))
+        {
+            LOG_TRACE(log, "Setting nice to " << new_os_thread_priority);
+
+            if (0 != setpriority(PRIO_PROCESS, os_thread_id, new_os_thread_priority))
+                throwFromErrno("Cannot 'setpriority'", ErrorCodes::CANNOT_SET_THREAD_PRIORITY);
+
+            os_thread_priority = new_os_thread_priority;
+        }
+    }
+#endif
+
     initPerformanceCounters();
     thread_state = ThreadState::AttachedToQuery;
 }
@@ -143,6 +173,18 @@ void ThreadStatus::detachQuery(bool exit_if_already_detached, bool thread_exits)
     thread_group.reset();
 
     thread_state = thread_exits ? ThreadState::Died : ThreadState::DetachedFromQuery;
+
+#if defined(__linux__)
+    if (os_thread_priority)
+    {
+        LOG_TRACE(log, "Resetting nice");
+
+        if (0 != setpriority(PRIO_PROCESS, os_thread_id, 0))
+            LOG_ERROR(log, "Cannot 'setpriority' back to zero: " << errnoToString(ErrorCodes::CANNOT_SET_THREAD_PRIORITY, errno));
+
+        os_thread_priority = 0;
+    }
+#endif
 }
 
 void ThreadStatus::logToQueryThreadLog(QueryThreadLog & thread_log)

debian/clickhouse-server.postinst

@@ -121,9 +121,9 @@ Please fix this and reinstall this package." >&2
     TMPFILE=/tmp/test_setcap.sh
 
     command -v setcap >/dev/null \
-        && echo > $TMPFILE && chmod a+x $TMPFILE && $TMPFILE && setcap "cap_net_admin,cap_ipc_lock+ep" $TMPFILE && $TMPFILE && rm $TMPFILE \
-        && setcap "cap_net_admin,cap_ipc_lock+ep" "${CLICKHOUSE_BINDIR}/${CLICKHOUSE_GENERIC_PROGRAM}" \
-        || echo "Cannot set 'net_admin' or 'ipc_lock' capability for clickhouse binary. This is optional. Taskstats accounting will be disabled. To enable taskstats accounting you may add the required capability later manually."
+        && echo > $TMPFILE && chmod a+x $TMPFILE && $TMPFILE && setcap "cap_net_admin,cap_ipc_lock,cap_sys_nice+ep" $TMPFILE && $TMPFILE && rm $TMPFILE \
+        && setcap "cap_net_admin,cap_ipc_lock,cap_sys_nice+ep" "${CLICKHOUSE_BINDIR}/${CLICKHOUSE_GENERIC_PROGRAM}" \
+        || echo "Cannot set 'net_admin' or 'ipc_lock' or 'sys_nice' capability for clickhouse binary. This is optional. Taskstats accounting will be disabled. To enable taskstats accounting you may add the required capability later manually."
 
     # Clean old dynamic compilation results
     if [ -d "${CLICKHOUSE_DATADIR_FROM_CONFIG}/build" ]; then

debian/clickhouse-server.service

@@ -11,7 +11,7 @@ RuntimeDirectory=clickhouse-server
 ExecStart=/usr/bin/clickhouse-server --config=/etc/clickhouse-server/config.xml --pid-file=/run/clickhouse-server/clickhouse-server.pid
 LimitCORE=infinity
 LimitNOFILE=500000
-CapabilityBoundingSet=CAP_NET_ADMIN CAP_IPC_LOCK
+CapabilityBoundingSet=CAP_NET_ADMIN CAP_IPC_LOCK CAP_SYS_NICE
 
 [Install]
 WantedBy=multi-user.target