Add new metrics

.github/workflows/test.yml

@@ -18,15 +18,28 @@ jobs:
           - "2.5"
           - "2.6"
           - "2.7"
+          - "2.8"
         cartridge: [ "", "1.2.0", "2.1.2", "2.4.0", "2.5.1", "2.6.0", "2.7.3" ]
+        include:
+          - tarantool: "2.x-latest"
+            cartridge: "2.7.3"
+          - tarantool: "2.x-latest"
+            cartridge: ""
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v2
 
       - uses: tarantool/setup-tarantool@v1
+        if: matrix.tarantool != '2.x-latest'
         with:
           tarantool-version: ${{ matrix.tarantool }}
 
+      - name: Install latest pre-release Tarantool 2.x
+        if: matrix.tarantool == '2.x-latest'
+        run: |
+          curl -L https://tarantool.io/pre-release/2/installer.sh | bash
+          sudo apt-get -y install tarantool
+
       - name: lint
         run: make lint
         env:

CHANGELOG.md

@@ -28,6 +28,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   - `tnt_cpu_number` (same as `tnt_cpu_count`)
   - `tnt_cpu_time` (same as `tnt_cpu_total`)
   - `tnt_vinyl_scheduler_dump_total` (same as `tnt_vinyl_scheduler_dump_count`)
+  - `tnt_replication_lag`
+  - `tnt_vinyl_regulator_blocked_writers`
+  - `tnt_net_requests_in_progress_total`
+  - `tnt_net_requests_in_progress_current`
+  - `tnt_net_requests_in_stream_total`
+  - `tnt_net_requests_in_stream_current`
+  - `tnt_replication_lsn`
 
 ### Deprecated
 
@@ -43,6 +50,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   - `tnt_cpu_count`
   - `tnt_cpu_total`
   - `tnt_vinyl_scheduler_dump_count`
+  - `tnt_replication_<id>_lag`
+  - `tnt_replication_master_<id>_lsn`
+  - `tnt_replication_replica_<id>_lsn`
 
 ## [0.12.0] - 2021-11-18
 ### Changed

Makefile

@@ -4,7 +4,7 @@ rpm:
 
 .rocks: metrics-scm-1.rockspec
 	tarantoolctl rocks make
-	tarantoolctl rocks install luatest 0.5.6
+	tarantoolctl rocks install luatest 0.5.7
 	tarantoolctl rocks install luacov 0.13.0
 	tarantoolctl rocks install luacheck 0.26.0
 	if [ -z $(CARTRIDGE_VERSION) ]; then \

doc/monitoring/metrics_reference.rst

@@ -219,6 +219,33 @@ Requests:
         *   -   ``tnt_net_requests_current``
             -   Number of pending network requests
 
+Requests in progress:
+
+..  container:: table
+
+    ..  list-table::
+        :widths: 25 75
+        :header-rows: 0
+
+        *   -   ``tnt_net_requests_in_progress_total``
+            -   Total count of requests processed by tx thread
+        *   -   ``tnt_net_requests_in_progress_current``
+            -   Count of requests currently being processed in the tx thread
+
+Requests placed in queues of streams:
+
+..  container:: table
+
+    ..  list-table::
+        :widths: 25 75
+        :header-rows: 0
+
+        *   -   ``tnt_net_requests_in_stream_total``
+            -   Total count of requests, which was placed in queues of streams
+                for all time
+        *   -   ``tnt_net_requests_in_stream_current``
+            -   Count of requests currently waiting in queues of streams
+
 ..  _metrics-reference-fibers:
 
 Fibers
@@ -316,12 +343,16 @@ Learn more about :ref:`replication in Tarantool <replication-mechanism>`.
             -   LSN number in vclock.
                 This metric always has the label ``{id="id"}``,
                 where ``id`` is the instance's number in the replica set.
-        *   -   ``tnt_replication_replica_<id>_lsn`` / ``tnt_replication_master_<id>_lsn``
-            -   LSN of the master/replica, where
-                ``id`` is the instance's number in the replica set.
-        *   -   ``tnt_replication_<id>_lag``
-            -   Replication lag value in seconds, where
-                ``id`` is the instance's number in the replica set.
+        *   -   ``tnt_replication_lsn``
+            -   LSN of the tarantool instance.
+                This metric always has labels ``{id="id", type="type"}``, where
+                ``id`` is the instance's number in the replica set,
+                ``type`` is ``master`` or ``replica``.
+        *   -   ``tnt_replication_lag``
+            -   Replication lag value in seconds.
+                This metric always has labels ``{id="id", stream="stream"}``,
+                where ``id`` is the instance's number in the replica set,
+                ``stream`` is ``downstream`` or ``upstream``.
 
 ..  _metrics-reference-runtime:
 
@@ -556,6 +587,9 @@ efficient.
                 The value is slightly smaller
                 than the amount of memory allocated for vinyl trees,
                 reflected in the :ref:`vinyl_memory <cfg_storage-vinyl_memory>` parameter.
+        *   -   ``tnt_vinyl_regulator_blocked_writers``
+            -   The number of fibers that are blocked waiting
+                for Vinyl level0 memory quota.
 
 Transactional activity
 ~~~~~~~~~~~~~~~~~~~~~~

metrics/init.lua

@@ -126,7 +126,8 @@ return {
     enable_default_metrics = function(include, exclude)
         log.warn('metrics tnt_net_sent_rps, tnt_net_received_rps, tnt_net_connections_rps, '..
             'tnt_net_requests_rps, tnt_stats_op_rps, tnt_space_count, tnt_fiber_count, ' ..
-            'lj_gc_total, tnt_cpu_count, tnt_cpu_total, tnt_vinyl_scheduler_dump_count ' ..
+            'lj_gc_total, tnt_cpu_count, tnt_cpu_total, tnt_vinyl_scheduler_dump_count, ' ..
+            'tnt_replication_<id>_lag, tnt_replication_master/replica_<id>_lsn ' ..
             'are deprecated and will be removed in next releases.')
         require('metrics.tarantool').enable(include, exclude)
     end,

metrics/tarantool/info.lua

@@ -23,9 +23,15 @@ local function update_info_metrics()
 
     for k, v in ipairs(info.replication) do
         if v.upstream ~= nil then
-            local metric_name = 'replication_' .. k .. '_lag'
-            collectors_list[metric_name] =
-                utils.set_gauge(metric_name, 'Replication lag for instance ' .. k, v.upstream.lag)
+            local metric_name_old = 'replication_' .. k .. '_lag'
+            collectors_list[metric_name_old] =
+                utils.set_gauge(metric_name_old, 'Replication lag for instance ' .. k, v.upstream.lag)
+            collectors_list.replication_lag =
+                utils.set_gauge('replication_lag', 'Replication lag', v.upstream.lag, {stream = 'upstream', id = k})
+        end
+        if v.downstream ~= nil then
+            collectors_list.replication_lag =
+                utils.set_gauge('replication_lag', 'Replication lag', v.downstream.lag, {stream = 'downstream', id = k})
         end
     end
 

metrics/tarantool/network.lua

@@ -41,6 +41,25 @@ local function update_network_metrics()
         collectors_list.net_requests_current =
             utils.set_gauge('net_requests_current', 'Pending requests', box_stat_net.REQUESTS.current)
     end
+
+    if box_stat_net.REQUESTS_IN_PROGRESS ~= nil then
+        collectors_list.net_requests_in_progress_total =
+            utils.set_counter('net_requests_in_progress_total', 'Requests in progress total amount',
+            box_stat_net.REQUESTS_IN_PROGRESS.total)
+        collectors_list.net_requests_in_progress_current =
+            utils.set_gauge('net_requests_in_progress_current',
+            'Count of requests currently being processed in the tx thread', box_stat_net.REQUESTS_IN_PROGRESS.current)
+    end
+
+    if box_stat_net.REQUESTS_IN_STREAM_QUEUE ~= nil then
+        collectors_list.net_requests_in_stream_total =
+            utils.set_counter('net_requests_in_stream_queue_total',
+            'Total count of requests, which was placed in queues of streams',
+            box_stat_net.REQUESTS_IN_STREAM_QUEUE.total)
+        collectors_list.net_requests_in_stream_current =
+            utils.set_gauge('net_requests_in_stream_queue_current',
+            'count of requests currently waiting in queues of streams', box_stat_net.REQUESTS_IN_STREAM_QUEUE.current)
+    end
 end
 
 

metrics/tarantool/replicas.lua

@@ -16,6 +16,9 @@ local function update_replicas_metrics()
                 local lsn = replication_info.lsn
                 local metric_name = 'replication_replica_' .. k .. '_lsn'
                 collectors_list[metric_name] = utils.set_gauge(metric_name, 'lsn for replica ' .. k, lsn - v)
+
+                collectors_list.replication_lsn =
+                    utils.set_gauge('replication_lsn', 'lsn for instance', lsn - v, {type = 'replica', id = k})
             end
         end
     else
@@ -29,6 +32,12 @@ local function update_replicas_metrics()
                         'lsn for master ' .. k,
                         current_box_info.lsn - lsn
                     )
+                    collectors_list.replication_lsn = utils.set_gauge(
+                        'replication_lsn',
+                        'lsn for instance',
+                        current_box_info.lsn - lsn,
+                        {type = 'master', id = k}
+                    )
                 end
             end
         end

metrics/tarantool/vinyl.lua

@@ -24,6 +24,11 @@ local function update()
     collectors_list.vinyl_regulator_dump_watermark =
         utils.set_gauge('vinyl_regulator_dump_watermark', 'Point when dumping must occur',
         vinyl_stat.regulator.dump_watermark)
+    if vinyl_stat.regulator.blocked_writers ~= nil then
+        collectors_list.vinyl_regulator_blocked_writers =
+            utils.set_gauge('vinyl_regulator_blocked_writers', 'The number of fibers that are blocked waiting ' ..
+            'for Vinyl level0 memory quota', vinyl_stat.regulator.blocked_writers)
+    end
 
     collectors_list.vinyl_tx_conflict =
         utils.set_gauge('vinyl_tx_conflict', 'Count of transaction conflicts', vinyl_stat.tx.conflict)

test/tarantool/vinyl_test.lua

@@ -26,5 +26,10 @@ g.test_vinyl_metrics_present = function()
     local metrics_cnt = fun.iter(metrics.collect()):filter(function(x)
         return x.metric_name:find('tnt_vinyl')
     end):length()
-    t.assert_equals(metrics_cnt, 20)
+    if utils.is_version_less(_TARANTOOL, '2.8.3')
+    and utils.is_version_greater(_TARANTOOL, '2.0.0') then
+        t.assert_equals(metrics_cnt, 20)
+    else
+        t.assert_equals(metrics_cnt, 21)
+    end
 end

test/utils.lua

@@ -62,9 +62,13 @@ function utils.find_metric(metric_name, metrics_data)
     return #m > 0 and m or nil
 end
 
+local function to_number_multiple(...)
+    return unpack(fun.map(tonumber, {...}):totable())
+end
+
 function utils.is_version_less(ver_str, reference_ver_str)
-    local major, minor, patch = string.match(ver_str, '^(%d+).(%d+).(%d+)')
-    local ref_major, ref_minor, ref_patch = string.match(reference_ver_str, '^(%d+).(%d+).(%d+)')
+    local major, minor, patch = to_number_multiple(string.match(ver_str, '^(%d+).(%d+).(%d+)'))
+    local ref_major, ref_minor, ref_patch = to_number_multiple(string.match(reference_ver_str, '^(%d+).(%d+).(%d+)'))
 
     if ( major < ref_major ) or ( major == ref_major and minor < ref_minor) or
       ( major == ref_major and minor == ref_minor and patch < ref_patch) then
@@ -75,8 +79,8 @@ function utils.is_version_less(ver_str, reference_ver_str)
 end
 
 function utils.is_version_greater(ver_str, reference_ver_str)
-    local major, minor, patch = string.match(ver_str, '^(%d+).(%d+).(%d+)')
-    local ref_major, ref_minor, ref_patch = string.match(reference_ver_str, '^(%d+).(%d+).(%d+)')
+    local major, minor, patch = to_number_multiple(string.match(ver_str, '^(%d+).(%d+).(%d+)'))
+    local ref_major, ref_minor, ref_patch = to_number_multiple(string.match(reference_ver_str, '^(%d+).(%d+).(%d+)'))
 
     if ( major > ref_major ) or ( major == ref_major and minor > ref_minor) or
       ( major == ref_major and minor == ref_minor and patch > ref_patch) then