Feat: Switch from Prometheus summary to histogram vectors

Replaced Prometheus summary vectors with histogram vectors for HTTP request metrics. Updated related constants, function names, and tests to reflect the change. This allows for better distribution tracking and more accurate performance analysis.

Signed-off-by: Christian Roessner <[email protected]>

Author: Christian Roessner

server/global/const.go

@@ -1230,8 +1230,14 @@ const (
 	// LuaFnCreateCounterVec represents the identifier for creating a Prometheus CounterVec.
 	LuaFnCreateCounterVec = "create_counter_vec"
 
-	// LuaFnStartTimer specifies the identifier for starting a Prometheus timer.
-	LuaFnStartTimer = "start_timer"
+	// LuaFnCreateHistogramVec is a constant representing the Lua function name for creating a HistogramVec in Prometheus.
+	LuaFnCreateHistogramVec = "create_histogram_vec"
+
+	// LuaFnStartSummaryTimer specifies the identifier for starting a Prometheus timer.
+	LuaFnStartSummaryTimer = "start_summary_timer"
+
+	// LuaFnStartHistogramTimer represents the function name for starting a histogram timer in Lua scripts.
+	LuaFnStartHistogramTimer = "start_histogram_timer"
 
 	// LuaFnStopTimer defines the identifier for stopping a Prometheus timer.
 	LuaFnStopTimer = "stop_timer"

server/lua-plugins.d/actions/haveibeenpwnd.lua

@@ -86,9 +86,9 @@ function nauthilus_call_action(request)
         end
 
 
-        nauthilus_prometheus.create_summary_vec(N .. "_duration_seconds", "HTTP request to the haveibeenpwnd network", {"http"})
+        nauthilus_prometheus.create_histogram_vec(N .. "_duration_seconds", "HTTP request to the haveibeenpwnd network", {"http"})
 
-        local timer = nauthilus_prometheus.start_timer(N .. "_duration_seconds", {http="get"})
+        local timer = nauthilus_prometheus.start_histogram_timer(N .. "_duration_seconds", {http="get"})
         local result, err = http.get("https://api.pwnedpasswords.com/range/" .. hash:sub(1, 5), {
             timeout = "10s",
             headers = {

server/lua-plugins.d/actions/telegram.lua

@@ -151,9 +151,9 @@ function nauthilus_call_action(request)
         values.username = username
         values.pwnd_info = pwnd_info
 
-        nauthilus_prometheus.create_summary_vec(N .. "_duration_seconds", "HTTP request to the telegram network", {"bot"})
+        nauthilus_prometheus.create_histogram_vec(N .. "_duration_seconds", "HTTP request to the telegram network", {"bot"})
 
-        local timer = nauthilus_prometheus.start_timer(N .. "_duration_seconds", {bot="send"})
+        local timer = nauthilus_prometheus.start_histogram_timer(N .. "_duration_seconds", {bot="send"})
         local _, err_bat = bot:sendMessage({
             chat_id = tonumber(os.getenv("TELEGRAM_CHAT_ID")),
             text = headline .. mustache:render(":\n\nSESSION {{session}}\nTS {{ts}}\nIP {{client_ip}}\nHOSTNAME {{hostname}}\nPROTOCOL {{proto}}\nDISPLAY_NAME {{display_name}}\nACCOUNT {{account}}\nUNIQUE ID {{unique_user_id}}\nUSERNAME {{username}}\nPWND INFO {{pwnd_info}}", values)

server/lua-plugins.d/features/blocklist.lua

@@ -47,9 +47,9 @@ function nauthilus_call_feature(request)
     local payload, json_encode_err = json.encode(t)
     nauthilus_util.if_error_raise(json_encode_err)
 
-    nauthilus_prometheus.create_summary_vec(N .. "_duration_seconds", "HTTP request to the blocklist service", {"http"})
+    nauthilus_prometheus.create_histogram_vec(N .. "_duration_seconds", "HTTP request to the blocklist service", {"http"})
 
-    local timer = nauthilus_prometheus.start_timer(N .. "_duration_seconds", {http="post"})
+    local timer = nauthilus_prometheus.start_histogram_timer(N .. "_duration_seconds", {http="post"})
     local result, request_err = http.post(os.getenv("BLOCKLIST_URL"), {
         timeout = "10s",
         headers = {

server/lua-plugins.d/filters/geoip.lua

@@ -78,7 +78,7 @@ function nauthilus_call_filter(request)
         local payload, json_encode_err = json.encode(t)
         nauthilus_util.if_error_raise(json_encode_err)
 
-        nauthilus_prometheus.create_summary_vec(N .. "_duration_seconds", "HTTP request to the geoip-policyd service", {
+        nauthilus_prometheus.create_histogram_vec(N .. "_duration_seconds", "HTTP request to the geoip-policyd service", {
             "http",
         })
 
@@ -87,7 +87,7 @@ function nauthilus_call_filter(request)
             "status",
         })
 
-        local timer = nauthilus_prometheus.start_timer(N .. "_duration_seconds", {http="post"})
+        local timer = nauthilus_prometheus.start_histogram_timer(N .. "_duration_seconds", {http="post"})
         local  result, request_err = http.post(os.getenv("GEOIP_POLICY_URL"), {
             timeout = "10s",
             headers = {

server/lualib/metrics/statistics.go

@@ -7,8 +7,9 @@ import (
 )
 
 var (
-	summaries = make(map[string]*prometheus.SummaryVec)
-	counters  = make(map[string]*prometheus.CounterVec)
+	summaries  = make(map[string]*prometheus.SummaryVec)
+	counters   = make(map[string]*prometheus.CounterVec)
+	histograms = make(map[string]*prometheus.HistogramVec)
 )
 
 // createSummaryVec registers a new Prometheus SummaryVec metric with the provided name, help description, and label names.
@@ -73,8 +74,39 @@ func createCounterVec(L *lua.LState) int {
 	return 0
 }
 
-// startTimer starts a Prometheus timer for a specified SummaryVec metric with the provided label values.
-func startTimer(L *lua.LState) int {
+// createHistogramVec registers a new Prometheus HistogramVec with the specified name, help message, and optional label names.
+func createHistogramVec(L *lua.LState) int {
+	name := L.CheckString(1)
+	help := L.CheckString(2)
+
+	labelNames := make([]string, 0)
+
+	if L.GetTop() > 2 {
+		labelTable := L.CheckTable(3)
+		labelTable.ForEach(func(_ lua.LValue, value lua.LValue) {
+			labelNames = append(labelNames, value.String())
+		})
+	}
+
+	// Check if the histogram already exists
+	if _, exists := histograms[name]; exists {
+		return 0
+	}
+
+	histogram := prometheus.NewHistogramVec(prometheus.HistogramOpts{
+		Name: name,
+		Help: help,
+	}, labelNames)
+
+	prometheus.MustRegister(histogram)
+
+	histograms[name] = histogram
+
+	return 0
+}
+
+// startSumaryTimer starts a Prometheus timer for a specified SummaryVec metric with the provided label values.
+func startSumaryTimer(L *lua.LState) int {
 	name := L.CheckString(1)
 	labels := L.CheckTable(2)
 
@@ -99,6 +131,32 @@ func startTimer(L *lua.LState) int {
 	return 1
 }
 
+// startHistogramTimer starts a timer for a Prometheus histogram with given name and labels.
+func startHistogramTimer(L *lua.LState) int {
+	name := L.CheckString(1)
+	labels := L.CheckTable(2)
+
+	histogram, exists := histograms[name]
+	if !exists {
+		L.ArgError(1, "HistogramVec not found")
+
+		return 0
+	}
+
+	labelValues := make(map[string]string)
+	labels.ForEach(func(key, value lua.LValue) {
+		labelValues[key.String()] = value.String()
+	})
+
+	timer := prometheus.NewTimer(histogram.With(labelValues))
+	ud := L.NewUserData()
+	ud.Value = timer
+
+	L.Push(ud)
+
+	return 1
+}
+
 // stopTimer stops a running Prometheus timer, recording its duration in the underlying metric.
 func stopTimer(L *lua.LState) int {
 	ud := L.CheckUserData(1)
@@ -132,13 +190,18 @@ func incrementCounter(L *lua.LState) int {
 }
 
 var exportsModPrometheus = map[string]lua.LGFunction{
-	global.LuaFnCreateSummaryVec: createSummaryVec,
-	global.LuaFnCreateCounterVec: createCounterVec,
-	global.LuaFnStartTimer:       startTimer,
-	global.LuaFnStopTimer:        stopTimer,
-	global.LuaFnIncrementCounter: incrementCounter,
+	global.LuaFnCreateSummaryVec:    createSummaryVec,
+	global.LuaFnCreateCounterVec:    createCounterVec,
+	global.LuaFnCreateHistogramVec:  createHistogramVec,
+	global.LuaFnStartSummaryTimer:   startSumaryTimer,
+	global.LuaFnStartHistogramTimer: startHistogramTimer,
+	global.LuaFnStopTimer:           stopTimer,
+	global.LuaFnIncrementCounter:    incrementCounter,
 }
 
+// LoaderModPrometheus loads the Prometheus module into the given Lua state.
+// It sets up the module's functions and pushes the module onto the stack.
+// Returns 1 to indicate the number of return values for the Lua stack.
 func LoaderModPrometheus(L *lua.LState) int {
 	mod := L.SetFuncs(L.NewTable(), exportsModPrometheus)
 

server/lualib/metrics/statistics_test.go

@@ -28,7 +28,7 @@ func TestCreateAndUseSummaryVec(t *testing.T) {
 		prometheus.create_summary_vec("test_summary", "Test Summary", {"label1", "label2"})
 
 		-- Start timer
-		timer = prometheus.start_timer("test_summary", {label1="value1", label2="value2"})
+		timer = prometheus.start_summary_timer("test_summary", {label1="value1", label2="value2"})
 
 		-- Some operation...
 
@@ -80,3 +80,40 @@ func TestCreateAndUseCounterVec(t *testing.T) {
 	value := testutil.ToFloat64(counter.WithLabelValues("value1", "value2"))
 	assert.Equal(t, float64(1), value, "Counter value should be 1")
 }
+
+func TestCreateAndUseHistogramVec(t *testing.T) {
+	L := lua.NewState()
+
+	defer L.Close()
+
+	// Register the module
+	L.PreloadModule("prometheus", LoaderModPrometheus)
+
+	err := runLuaCode(L, `
+		local prometheus = require("prometheus")
+		
+		-- Create a HistogramVec
+		prometheus.create_histogram_vec("test_histogram", "Histogram test", {"label1", "label2"})
+		
+		-- Start timer
+		timer = prometheus.start_histogram_timer("test_histogram", {label1 = "value1", label2 = "value2"})
+		
+		-- Some operation...
+		
+		-- Stop timer
+		prometheus.stop_timer(timer)
+	`)
+
+	if err != nil {
+		t.Fatalf("Lua code execution failed: %v", err)
+	}
+
+	// Verify the histogram was created and contains the recorded value.
+	histogram, exists := histograms["test_histogram"]
+
+	assert.True(t, exists, "HistogramVec 'test_histogram' should exist")
+
+	// Test whether the histogram has recorded data
+	count := testutil.CollectAndCount(histogram)
+	assert.NotZero(t, count, "Expected non-zero count in HistogramVec")
+}

server/stats/statistics.go

@@ -119,7 +119,7 @@ var (
 	})
 
 	// FunctionDuration variable declaration that creates a new Prometheus SummaryVec with the specified name and help message, and with "service" and "method" labels.
-	FunctionDuration = promauto.NewSummaryVec(prometheus.SummaryOpts{
+	FunctionDuration = promauto.NewHistogramVec(prometheus.HistogramOpts{
 		Name: "function_duration_seconds",
 		Help: "Time spent in function",
 	}, []string{"service", "task"})