prosody-hg: core/statsmanager.lua comparison

comparison core/statsmanager.lua @ 11523:5f15ab7c6ae5

Statistics: Rewrite statistics backends to use OpenMetrics The metric subsystem of Prosody has had some shortcomings from the perspective of the current state-of-the-art in metric observability. The OpenMetrics standard [0] is a formalization of the data model (and serialization format) of the well-known and widely-used Prometheus [1] software stack. The previous stats subsystem of Prosody did not map well to that format (see e.g. [2] and [3]); the key reason is that it was trying to do too much math on its own ([2]) while lacking first-class support for "families" of metrics ([3]) and structured metric metadata (despite the `extra` argument to metrics, there was no standard way of representing common things like "tags" or "labels"). Even though OpenMetrics has grown from the Prometheus world of monitoring, it maps well to other popular monitoring stacks such as: - InfluxDB (labels can be mapped to tags and fields as necessary) - Carbon/Graphite (labels can be attached to the metric name with dot-separation) - StatsD (see graphite when assuming that graphite is used as backend, which is the default) The util.statsd module has been ported to use the OpenMetrics model as a proof of concept. An implementation which exposes the util.statistics backend data as Prometheus metrics is ready for publishing in prosody-modules (most likely as mod_openmetrics_prometheus to avoid breaking existing 0.11 deployments). At the same time, the previous measure()-based API had one major advantage: It is really simple and easy to use without requiring lots of knowledge about OpenMetrics or similar concepts. For that reason as well as compatibility with existing code, it is preserved and may even be extended in the future. However, code relying on the `stats-updated` event as well as `get_stats` from `statsmanager` will break because the data model has changed completely; in case of `stats-updated`, the code will simply not run (as the event was renamed in order to avoid conflicts); the `get_stats` function has been removed completely (so it will cause a traceback when it is attempted to be used). Note that the measure_*_event methods have been removed from the module API. I was unable to find any uses or documentation and thus deemed they should not be ported. Re-implementation is possible when necessary. [0]: https://openmetrics.io/ [1]: https://prometheus.io/ [2]: #959 [3]: #960

author	Jonas Schäfer <jonas@wielicki.name>
date	Sun, 18 Apr 2021 11:47:41 +0200
parents	10d13e0554f9
children	76d32b2ca5eb

comparison

equal deleted inserted replaced

-:5bd38d9197e1
+:5f15ab7c6ae5
 local config = require "core.configmanager";
 local log = require "util.logger".init("stats");
 local timer = require "util.timer";
 local fire_event = prosody.events.fire_event;
+local array = require "util.array";
+local timed = require "util.openmetrics".timed;
 local stats_interval_config = config.get("*", "statistics_interval");
 local stats_interval = tonumber(stats_interval_config);
 if stats_interval_config and not stats_interval and stats_interval_config ~= "manual" then
 	log("error", "Invalid 'statistics_interval' setting, statistics will be disabled");
 if stats == nil then
 	log("error", "Error loading statistics provider '%s': %s", stats_provider, stats_err);
 end
-local measure, collect;
+local measure, collect, metric, cork, uncork;
-local latest_stats = {};
-local changed_stats = {};
-local stats_extra = {};
 if stats then
-	function measure(type, name, conf)
+	function metric(type_, name, unit, description, labels, extra)
-		local f = assert(stats[type], "unknown stat type: "..type);
+		local registry = stats.metric_registry
-		return f(name, conf);
+		local f = assert(registry[type_], "unknown metric family type: "..type_);
+		return f(registry, name, unit or "", description or "", labels, extra);
+	end
+	local function new_legacy_metric(stat_type, name, unit, description, fixed_label_key, fixed_label_value, extra)
+		local label_keys = array()
+		local conf = extra or {}
+		if fixed_label_key then
+			label_keys:push(fixed_label_key)
+		end
+		unit = unit or ""
+		local mf = metric(stat_type, "prosody_" .. name, unit, description, label_keys, conf);
+		if fixed_label_key then
+			mf = mf:with_partial_label(fixed_label_value)
+		end
+		return mf:with_labels()
+	end
+	local function unwrap_legacy_extra(extra, type_, name, unit)
+		local description = extra and extra.description or "Legacy "..type_.." metric "..name
+		unit = extra and extra.unit or unit
+		return description, unit
+	end
+	-- These wrappers provide the pre-OpenMetrics interface of statsmanager
+	-- and moduleapi (module:measure).
+	local legacy_metric_wrappers = {
+		amount = function(name, fixed_label_key, fixed_label_value, extra)
+			local initial = 0
+			if type(extra) == "number" then
+				initial = extra
+			else
+				initial = extra and extra.initial or initial
+			end
+			local description, unit = unwrap_legacy_extra(extra, "amount", name)
+			local m = new_legacy_metric("gauge", name, unit, description, fixed_label_key, fixed_label_value)
+			m:set(initial or 0)
+			return function(v)
+				m:set(v)
+			end
+		end;
+		counter = function(name, fixed_label_key, fixed_label_value, extra)
+			if type(extra) == "number" then
+				-- previous versions of the API allowed passing an initial
+				-- value here; we do not allow that anymore, it is not a thing
+				-- which makes sense with counters
+				extra = nil
+			end
+			local description, unit = unwrap_legacy_extra(extra, "counter", name)
+			local m = new_legacy_metric("counter", name, unit, description, fixed_label_key, fixed_label_value)
+			m:set(0)
+			return function(v)
+				m:add(v)
+			end
+		end;
+		rate = function(name, fixed_label_key, fixed_label_value, extra)
+			if type(extra) == "number" then
+				-- previous versions of the API allowed passing an initial
+				-- value here; we do not allow that anymore, it is not a thing
+				-- which makes sense with counters
+				extra = nil
+			end
+			local description, unit = unwrap_legacy_extra(extra, "counter", name)
+			local m = new_legacy_metric("counter", name, unit, description, fixed_label_key, fixed_label_value)
+			m:set(0)
+			return function()
+				m:add(1)
+			end
+		end;
+		times = function(name, fixed_label_key, fixed_label_value, extra)
+			local conf = {}
+			if extra and extra.buckets then
+				conf.buckets = extra.buckets
+			else
+				conf.buckets = { 0.001, 0.01, 0.1, 1.0, 10.0, 100.0 }
+			end
+			local description, _ = unwrap_legacy_extra(extra, "times", name)
+			local m = new_legacy_metric("histogram", name, "seconds", description, fixed_label_key, fixed_label_value, conf)
+			return function()
+				return timed(m)
+			end
+		end;
+		sizes = function(name, fixed_label_key, fixed_label_value, extra)
+			local conf = {}
+			if extra and extra.buckets then
+				conf.buckets = extra.buckets
+			else
+				conf.buckets = { 1024, 4096, 32768, 131072, 1048576, 4194304, 33554432, 134217728, 1073741824 }
+			end
+			local description, _ = unwrap_legacy_extra(extra, "sizes", name)
+			local m = new_legacy_metric("histogram", name, "bytes", description, fixed_label_key, fixed_label_value, conf)
+			return function(v)
+				m:sample(v)
+			end
+		end;
+		distribution = function(name, fixed_label_key, fixed_label_value, extra)
+			if type(extra) == "string" then
+				-- compat with previous API
+				extra = { unit = extra }
+			end
+			local description, unit = unwrap_legacy_extra(extra, "distribution", name, "")
+			local m = new_legacy_metric("summary", name, unit, description, fixed_label_key, fixed_label_value)
+			return function(v)
+				m:sample(v)
+			end
+		end;
+	};
+	-- Argument order switched here to support the legacy statsmanager.measure
+	-- interface.
+	function measure(stat_type, name, extra, fixed_label_key, fixed_label_value)
+		local wrapper = assert(legacy_metric_wrappers[stat_type], "unknown legacy metric type "..stat_type)
+		return wrapper(name, fixed_label_key, fixed_label_value, extra)
+	end
+	if stats.cork then
+		function cork()
+			return stats:cork()
+		end
+		function uncork()
+			return stats:uncork()
+		end
+	else
+		function cork() end
+		function uncork() end
 	end
 	if stats_interval or stats_interval_config == "manual" then
 		local mark_collection_start = measure("times", "stats.collection");
 		local mark_processing_start = measure("times", "stats.processing");
 		function collect()
 			local mark_collection_done = mark_collection_start();
 			fire_event("stats-update");
+			-- ensure that the backend is uncorked, in case it got stuck at
+			-- some point, to avoid infinite resource use
+			uncork()
 			mark_collection_done();
+			local manual_result = nil
-			if stats.get_stats then
+			if stats.metric_registry then
+				-- only if supported by the backend, we fire the event which
+				-- provides the current metric values
 				local mark_processing_done = mark_processing_start();
-				changed_stats, stats_extra = {}, {};
+				local metric_registry = stats.metric_registry;
-				for stat_name, getter in pairs(stats.get_stats()) do
+				fire_event("openmetrics-updated", { metric_registry = metric_registry })
-					-- luacheck: ignore 211/type
-					local type, value, extra = getter();
-					local old_value = latest_stats[stat_name];
-					latest_stats[stat_name] = value;
-					if value ~= old_value then
-						changed_stats[stat_name] = value;
-					end
-					if extra then
-						stats_extra[stat_name] = extra;
-					end
-				end
-				fire_event("stats-updated", { stats = latest_stats, changed_stats = changed_stats, stats_extra = stats_extra });
 				mark_processing_done();
-			end
+				manual_result = metric_registry;
-			return stats_interval;
+			end
+			return stats_interval, manual_result;
 		end
 		if stats_interval then
 			log("debug", "Statistics enabled using %s provider, collecting every %d seconds", stats_provider_name, stats_interval);
 			timer.add_task(stats_interval, collect);
 			prosody.events.add_handler("server-started", function () collect() end, -1);
 		log("debug", "Statistics enabled using %s provider, collection is disabled", stats_provider_name);
 	end
 else
 	log("debug", "Statistics disabled");
 	function measure() return measure; end
+	local dummy_mt = {}
+	function dummy_mt.__newindex()
+	end
+	function dummy_mt:__index()
+		return self
+	end
+	function dummy_mt:__call()
+		return self
+	end
+	local dummy = {}
+	setmetatable(dummy, dummy_mt)
+	function metric() return dummy; end
+	function cork() end
+	function uncork() end
 end
 local exported_collect = nil;
 if stats_interval_config == "manual" then
 	exported_collect = collect;
 end
 return {
 	collect = exported_collect;
 	measure = measure;
-	get_stats = function ()
+	cork = cork;
-		return latest_stats, changed_stats, stats_extra;
+	uncork = uncork;
-	end;
+	metric = metric;
-	get = function (name)
+	get_metric_registry = function ()
-		return latest_stats[name], stats_extra[name];
+		return stats and stats.metric_registry or nil
 	end;
 };

Mercurial > prosody-hg

comparison core/statsmanager.lua @ 11523:5f15ab7c6ae5