File

util/statsd.lua @ 13652:a08065207ef0

net.server_epoll: Call :shutdown() on TLS sockets when supported Comment from Matthew: This fixes a potential issue where the Prosody process gets blocked on sockets waiting for them to close. Unlike non-TLS sockets, closing a TLS socket sends layer 7 data, and this can cause problems for sockets which are in the process of being cleaned up. This depends on LuaSec changes which are not yet upstream. From Martijn's original email: So first my analysis of luasec. in ssl.c the socket is put into blocking mode right before calling SSL_shutdown() inside meth_destroy(). My best guess to why this is is because meth_destroy is linked to the __close and __gc methods, which can't exactly be called multiple times and luasec does want to make sure that a tls session is shutdown as clean as possible. I can't say I disagree with this reasoning and don't want to change this behaviour. My solution to this without changing the current behaviour is to introduce a shutdown() method. I am aware that this overlaps in a conflicting way with tcp's shutdown method, but it stays close to the OpenSSL name. This method calls SSL_shutdown() in the current (non)blocking mode of the underlying socket and returns a boolean whether or not the shutdown is completed (matching SSL_shutdown()'s 0 or 1 return values), and returns the familiar ssl_ioerror() strings on error with a false for completion. This error can then be used to determine if we have wantread/wantwrite to finalize things. Once meth_shutdown() has been called once a shutdown flag will be set, which indicates to meth_destroy() that the SSL_shutdown() has been handled by the application and it shouldn't be needed to set the socket to blocking mode. I've left the SSL_shutdown() call in the LSEC_STATE_CONNECTED to prevent TOCTOU if the application reaches a timeout for the shutdown code, which might allow SSL_shutdown() to clean up anyway at the last possible moment. Another thing I've changed to luasec is the call to socket_setblocking() right before calling close(2) in socket_destroy() in usocket.c. According to the latest POSIX[0]: Note that the requirement for close() on a socket to block for up to the current linger interval is not conditional on the O_NONBLOCK setting. Which I read to mean that removing O_NONBLOCK on the socket before close doesn't impact the behaviour and only causes noise in system call tracers. I didn't touch the windows bits of this, since I don't do windows. For the prosody side of things I've made the TLS shutdown bits resemble interface:onwritable(), and put it under a combined guard of self._tls and self.conn.shutdown. The self._tls bit is there to prevent getting stuck on this condition, and self.conn.shutdown is there to prevent the code being called by instances where the patched luasec isn't deployed. The destroy() method can be called from various places and is read by me as the "we give up" error path. To accommodate for these unexpected entrypoints I've added a single call to self.conn:shutdown() to prevent the socket being put into blocking mode. I have no expectations that there is any other use here. Same as previous, the self.conn.shutdown check is there to make sure it's not called on unpatched luasec deployments and self._tls is there to make sure we don't call shutdown() on tcp sockets. I wouldn't recommend logging of the conn:shutdown() error inside close(), since a lot of clients simply close the connection before SSL_shutdown() is done.
author Martijn van Duren <martijn@openbsd.org>
date Thu, 06 Feb 2025 15:04:38 +0000
parent 12975:d10957394a3c
line wrap: on
line source

local socket = require "socket";
local time = require "prosody.util.time".now;
local array = require "prosody.util.array";
local t_concat = table.concat;

local new_metric_registry = require "prosody.util.openmetrics".new_metric_registry;
local render_histogram_le = require "prosody.util.openmetrics".render_histogram_le;

-- BEGIN of Metric implementations

-- Gauges
local gauge_metric_mt = {}
gauge_metric_mt.__index = gauge_metric_mt

local function new_gauge_metric(full_name, impl)
	local metric = {
		_full_name = full_name;
		_impl = impl;
		value = 0;
	}
	setmetatable(metric, gauge_metric_mt)
	return metric
end

function gauge_metric_mt:set(value)
	self.value = value
	self._impl:push_gauge(self._full_name, value)
end

function gauge_metric_mt:add(delta)
	self.value = self.value + delta
	self._impl:push_gauge(self._full_name, self.value)
end

function gauge_metric_mt:reset()
	self.value = 0
	self._impl:push_gauge(self._full_name, 0)
end

function gauge_metric_mt.iter_samples()
	-- statsd backend does not support iteration.
	return function()
		return nil
	end
end

-- Counters
local counter_metric_mt = {}
counter_metric_mt.__index = counter_metric_mt

local function new_counter_metric(full_name, impl)
	local metric = {
		_full_name = full_name,
		_impl = impl,
		value = 0,
	}
	setmetatable(metric, counter_metric_mt)
	return metric
end

function counter_metric_mt:set(value)
	local delta = value - self.value
	self.value = value
	self._impl:push_counter_delta(self._full_name, delta)
end

function counter_metric_mt:add(value)
	self.value = (self.value or 0) + value
	self._impl:push_counter_delta(self._full_name, value)
end

function counter_metric_mt.iter_samples()
	-- statsd backend does not support iteration.
	return function()
		return nil
	end
end

function counter_metric_mt:reset()
	self.value = 0
end

-- Histograms
local histogram_metric_mt = {}
histogram_metric_mt.__index = histogram_metric_mt

local function new_histogram_metric(buckets, full_name, impl)
	-- NOTE: even though the more or less proprietary dogstatsd has Its own
	-- histogram implementation, we push the individual buckets in this statsd
	-- backend for both consistency and compatibility across statsd
	-- implementations.
	local metric = {
		_sum_name = full_name..".sum",
		_count_name = full_name..".count",
		_impl = impl,
		_created = time(),
		_sum = 0,
		_count = 0,
	}
	-- the order of buckets matters unfortunately, so we cannot directly use
	-- the thresholds as table keys
	for i, threshold in ipairs(buckets) do
		local threshold_s = render_histogram_le(threshold)
		metric[i] = {
			threshold = threshold,
			threshold_s = threshold_s,
			count = 0,
			_full_name = full_name..".bucket."..(threshold_s:gsub("%.", "_")),
		}
	end
	setmetatable(metric, histogram_metric_mt)
	return metric
end

function histogram_metric_mt:sample(value)
	-- According to the I-D, values must be part of all buckets
	for i, bucket in pairs(self) do
		if "number" == type(i) and value <= bucket.threshold then
			bucket.count = bucket.count + 1
			self._impl:push_counter_delta(bucket._full_name, 1)
		end
	end
	self._sum = self._sum + value
	self._count = self._count + 1
	self._impl:push_gauge(self._sum_name, self._sum)
	self._impl:push_counter_delta(self._count_name, 1)
end

function histogram_metric_mt.iter_samples()
	-- statsd backend does not support iteration.
	return function()
		return nil
	end
end

function histogram_metric_mt:reset()
	self._created = time()
	self._count = 0
	self._sum = 0
	for i, bucket in pairs(self) do
		if "number" == type(i) then
			bucket.count = 0
		end
	end
	self._impl:push_gauge(self._sum_name, self._sum)
end

-- Summaries
local summary_metric_mt = {}
summary_metric_mt.__index = summary_metric_mt

local function new_summary_metric(full_name, impl)
	local metric = {
		_sum_name = full_name..".sum",
		_count_name = full_name..".count",
		_impl = impl,
	}
	setmetatable(metric, summary_metric_mt)
	return metric
end

function summary_metric_mt:sample(value)
	self._impl:push_counter_delta(self._sum_name, value)
	self._impl:push_counter_delta(self._count_name, 1)
end

function summary_metric_mt.iter_samples()
	-- statsd backend does not support iteration.
	return function()
		return nil
	end
end

function summary_metric_mt.reset()
end

-- BEGIN of statsd client implementation

local statsd_mt = {}
statsd_mt.__index = statsd_mt

function statsd_mt:cork()
	self.corked = true
	self.cork_buffer = self.cork_buffer or {}
end

function statsd_mt:uncork()
	self.corked = false
	self:_flush_cork_buffer()
end

function statsd_mt:_flush_cork_buffer()
	local buffer = self.cork_buffer
	for metric_name, value in pairs(buffer) do
		self:_send_gauge(metric_name, value)
		buffer[metric_name] = nil
	end
end

function statsd_mt:push_gauge(metric_name, value)
	if self.corked then
		self.cork_buffer[metric_name] = value
	else
		self:_send_gauge(metric_name, value)
	end
end

function statsd_mt:_send_gauge(metric_name, value)
	self:_send(self.prefix..metric_name..":"..tostring(value).."|g")
end

function statsd_mt:push_counter_delta(metric_name, delta)
	self:_send(self.prefix..metric_name..":"..tostring(delta).."|c")
end

function statsd_mt:_send(s)
	return self.sock:send(s)
end

-- END of statsd client implementation

local function build_metric_name(family_name, labels)
	local parts = array { family_name }
	if labels then
		parts:append(labels)
	end
	return t_concat(parts, "/"):gsub("%.", "_"):gsub("/", ".")
end

local function new(config)
	if not config or not config.statsd_server then
		return nil, "No statsd server specified in the config, please see https://prosody.im/doc/statistics";
	end

	local sock = socket.udp();
	sock:setpeername(config.statsd_server, config.statsd_port or 8125);

	local prefix = (config.prefix or "prosody")..".";

	local impl = {
		metric_registry = nil;
		sock = sock;
		prefix = prefix;
	};
	setmetatable(impl, statsd_mt)

	local backend = {
		gauge = function(family_name, labels)
			return new_gauge_metric(build_metric_name(family_name, labels), impl)
		end;
		counter = function(family_name, labels)
			return new_counter_metric(build_metric_name(family_name, labels), impl)
		end;
		histogram = function(buckets, family_name, labels)
			return new_histogram_metric(buckets, build_metric_name(family_name, labels), impl)
		end;
		summary = function(family_name, labels, extra)
			return new_summary_metric(build_metric_name(family_name, labels), impl, extra)
		end;
	};

	impl.metric_registry = new_metric_registry(backend);

	return impl;
end

return {
	new = new;
}