cpu: add instance CPU usage metric

This patch adds new `tnt_cpu_instance` metric.

Closes #TNTP-4361

Author: patapenka-alexey

.github/workflows/test.yml

@@ -39,6 +39,8 @@ jobs:
             cartridge: ""
           - tarantool: "3.3"
             cartridge: ""
+          - tarantool: "3.4"
+            cartridge: ""
     runs-on: ubuntu-24.04
     steps:
       - uses: actions/checkout@v4

CHANGELOG.md

@@ -8,6 +8,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+- `tnt_cpu_instance` metric.
+
 ### Changed
 
 ### Fixed

doc/monitoring/metrics_reference.rst

@@ -527,6 +527,13 @@ They are only available on Linux.
                 *   ``kind`` can be either ``user`` or ``system``
                 *   ``thread_name`` is ``tarantool``, ``wal``, ``iproto``, or ``coio``
                 *   ``file_name`` is the entrypoint file name, for example, ``init.lua``.
+        *   -   ``tnt_cpu_instance``
+            -   Tarantool instance CPU time.
+                This metric always has the labels
+                ``{kind="user"}``,
+                where:
+
+                *   ``kind`` can be either ``user`` or ``system``.
 
 There are also two cross-platform metrics, which can be obtained with a ``getrusage()`` call.
 

metrics/psutils/cpu.lua

@@ -49,6 +49,20 @@ local function update_cpu_metrics()
         collectors_list.cpu_thread:remove(thread_info)
     end
     threads = new_threads
+
+    local instance_info = psutils.get_instance_cpu_time()
+    if instance_info ~= nil then
+        local instance_stime_labels = {kind = 'system'}
+        collectors_list.instance_cpu_usage = utils.set_gauge('cpu_instance',
+            'Tarantool instance cpu time', instance_info.stime,
+            instance_stime_labels, nil, {default = true})
+
+        local instance_utime_labels = {kind = 'user'}
+        collectors_list.instance_cpu_usage = utils.set_gauge('cpu_instance',
+            'Tarantool instance cpu time', instance_info.utime,
+            instance_utime_labels, nil, {default = true})
+
+    end
 end
 
 local function clear_cpu_metrics()

metrics/psutils/psutils_linux.lua

@@ -66,8 +66,13 @@ local function get_process_cpu_time()
     return thread_time
 end
 
+local function get_instance_cpu_time()
+    return parse_process_stat('/proc/self/stat')
+end
+
 return {
     get_cpu_time = get_cpu_time,
     get_process_cpu_time = get_process_cpu_time,
+    get_instance_cpu_time = get_instance_cpu_time,
     get_cpu_count = get_nprocs_conf,
 }

test/psutils_linux_test.lua

@@ -52,3 +52,11 @@ g.test_get_cpu_count = function(cg)
         t.assert_gt(psutils_linux.get_cpu_count(), 0)
     end)
 end
+
+g.test_get_instance_cpu_time = function(cg)
+    cg.server:exec(function()
+        local psutils_linux = require('metrics.psutils.psutils_linux')
+        local expected = {pid = 280908, comm = 'tarantool', utime = 222, stime = 111}
+        t.assert_items_equals(psutils_linux.get_instance_cpu_time(), expected)
+    end)
+end

test/psutils_linux_test_payload/init.lua

@@ -1,10 +1,12 @@
 local payload_dir = debug.sourcedir()
 local stat_file_path = payload_dir .. '/proc_stat'
+local self_stat_file_path = payload_dir .. '/proc_self_stat'
 local task_dir_path = payload_dir .. '/proc_self_task'
 
 return {
     files = {
         ['/proc/stat'] = stat_file_path,
+        ['/proc/self/stat'] = self_stat_file_path,
         ['/proc/self/task/1/stat'] = task_dir_path .. '/1/stat',
         ['/proc/self/task/12/stat'] = task_dir_path .. '/12/stat',
         ['/proc/self/task/13/stat'] = task_dir_path .. '/13/stat',

test/psutils_linux_test_payload/proc_self_stat

@@ -0,0 +1 @@
+280908 (tarantool) R 12484 280908 12484 34819 280908 4194304 101 0 0 0 222 111 0 0 20 0 1 0 50085946 8679424 448 18446744073709551615 110672370106368 110672370123953 140731302033040 0 0 0 0 0 0 0 0 0 17 9 0 0 0 0 0 110672370137744 110672370139240 110673314615296 140731302035060 140731302035080 140731302035080 140731302039531 0

test/psutils_linux_thread_clean_test.lua

@@ -56,6 +56,21 @@ g.test_cpu_count = function(cg)
     end)
 end
 
+g.test_instance_cpu = function(cg)
+    cg.server:exec(function()
+        local metrics = require('metrics')
+        local utils = require('test.utils') -- luacheck: ignore 431
+
+        require('metrics.psutils.cpu').update()
+
+        local observations = metrics.collect()
+        local metric = utils.find_metric('tnt_cpu_instance', observations)
+        t.assert_not_equals(metric, nil)
+        t.assert_equals(#metric, 2)
+        t.assert_gt(metric[1].value, 0)
+        t.assert_gt(metric[2].value, 0)
+    end)
+end
 
 g.test_clear = function(cg)
     cg.server:exec(function()
@@ -71,6 +86,7 @@ g.test_clear = function(cg)
                 'tnt_cpu_number',
                 'tnt_cpu_time',
                 'tnt_cpu_thread',
+                'tnt_cpu_instance',
             }
 
             for _, metric_name in ipairs(expected) do