postfix-exporter 2.0.3 → 2.1.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: d6e9b995c8c56b28d9500d51b20a8857b3fdcc44
-  data.tar.gz: 34c495ffd68b19de2c65c5297e58a7d197a0f705
+  metadata.gz: 59a89aac9e188bb3a7a5cf1617425bd797ebf68f
+  data.tar.gz: d7fd70f12daadfd79a90c66efce6c862afd31b00
 SHA512:
-  metadata.gz: 1483d066cc63929f244a0337f0c61ab2740d9d105fcc1af92881b79c0a5a819da42e81b64d82cccf5855853f3ef04342a4f09836d89e66fac545f48f0754e224
-  data.tar.gz: 8a1f46107f17a18d6d9ca6b64c2e6b6ad6e0a4b7f2e489495640af5ec82b8b08f371f741c31438b46fa39b935cf76c7304f98db1c58c9606a9369efac71052c2
+  metadata.gz: 0e51b58b74c279e81a3a88f9e3322f95f363571a6e959d14a8f3ebf1b3d77a5f899821f0673a9b347c1a9906574983bbc8ba297667917238a18eb933da55daed
+  data.tar.gz: 6e6ecffdb49e584e2592ba7d69bdc08fb67efd1abc4162024b7511a5e97a393ef6bc643ca8d44993744b0230d26d434b3eda8adff91d143b5c224ba3c12d9a02

data/README.md

@@ -7,8 +7,18 @@ server.
 * Examines mail queue periodically and exports `postfix_mail_queue_size`;
 
 * Reads syslog entries as they happen, and exports disposition status
-  counters (`postfix_disposition`) per DSN, as well as delay summaries per
-  DSN;
+  counters (`postfix_delivery_delays_count`) per DSN, as well as delay summaries per
+  DSN (`postfix_delivery_delays{quantile="..."}`);
+
+* Total number of SMTP connections (`postfix_smtpd_connections_total`) and
+  currently-active connections (`postfix_smtpd_active_connections`);
+
+* Count how many delivery attempts were received
+  (`postfix_incoming_delivery_attempts_total`), split out by whether we
+  accepted or rejected the message (`status`) and the exact DSN provided to
+  the client (`dsn`);
+
+* Whether or not the Postfix `master` process is running (`postfix_up`);
 
 * Drinks from the syslog stream directly.
 
@@ -21,6 +31,11 @@ available in the container as `/var/spool/postfix`), and some env
 vars.  You can also run it directly (via the gem), with the same env vars,
 and with the expectation that `/var/spool/postfix` is in the usual place.
 
+If you configure the postfix-exporter to run in the same PID namespace as
+whatever it is that's running Postfix itself (either `--pid=host` or in the
+same NS namespace, a la k8s pods), then the `postfix_up` metric will be
+valid, otherwise it'll be random (but unlikely to be correct).
+
 
 ## Environment Variables
 

data/bin/postfix-exporter

@@ -8,7 +8,11 @@ require 'logger'
 
 prometheus = Prometheus::Client.registry
 
+prometheus.gauge(:postfix_exporter_start_time_seconds, "When this process started up").set({}, Time.now.to_f)
+
 mailq = prometheus.gauge(:postfix_queue_size, "Number of messages in the mail queue")
+q_err = prometheus.counter(:postfix_queue_processing_error_total, "Exceptions raised whilst scanning the Postfix queue")
+up    = prometheus.gauge(:postfix_up, "Whether the master process is running or not")
 
 Thread.abort_on_exception = true
 
@@ -22,18 +26,49 @@ Thread.new do
 			# deferred is special, because it's often hueg it gets sharded into
 			# multiple subdirectories
 			mailq.set({ queue: 'deferred' }, Dir["/var/spool/postfix/deferred/*/*"].size)
-
-			sleep 5
 		rescue StandardError => ex
 			$stderr.puts "Error while monitoring queue sizes: #{ex.message} (#{ex.class})"
 			$stderr.puts ex.backtrace.map { |l| "  #{l}" }.join("\n")
-			sleep 1
+			q_err.increment(class: ex.class.to_s, phase: "scan")
 		end
+
+		begin
+			master_pid = File.read("/var/spool/postfix/pid/master.pid").to_i
+
+			if master_pid > 1
+				Process.kill(0, master_pid)
+				# If we get here, then the process exists, and
+				# that'll do for our purposes
+				up.set({}, 1)
+			else
+				up.set({}, 0)
+			end
+		rescue Errno::ENOENT, Errno::ESRCH, Errno::EACCES
+			up.set({}, 0)
+		rescue Errno::EPERM
+			# Ironically, we don't need to be able to *actually*
+			# signal the process; EPERM means it exists and is running
+			# as someone more privileged than us, which is enough
+			# for our purposes
+			up.set({}, 1)
+		rescue StandardError => ex
+			$stderr.puts "Error while checking master process: #{ex.message} (#{ex.class})"
+			$stderr.puts ex.backtrace.map { |l| "  #{l}" }.join("\n")
+			q_err.increment(class: ex.class.to_s, phase: "up")
+		end
+
+		sleep 5
+
 	end
 end
 
 if ENV["SYSLOG_SOCKET"]
-	delays   = prometheus.summary(:postfix_delivery_delays, "Distribution of time taken to deliver (or bounce) messages")
+	delays     = prometheus.summary(:postfix_delivery_delays, "Distribution of time taken to deliver (or bounce) messages")
+	connects   = prometheus.counter(:postfix_smtpd_connections_total, "Connections to smtpd")
+	active     = prometheus.gauge(:postfix_smtpd_active_connections, "Current connections to smtpd")
+	incoming   = prometheus.counter(:postfix_incoming_delivery_attempts_total, "Delivery attempts, labelled by dsn and status")
+	messages   = prometheus.counter(:postfix_log_messages_total, "Syslog messages received, labelled by how it was handled")
+	log_errors = prometheus.counter(:postfix_log_processing_error_total, "Exceptions raised whilst processing log messages")
 
 	Thread.new do
 		begin
@@ -55,10 +90,36 @@ if ENV["SYSLOG_SOCKET"]
 					if status == "bounced" or status == "sent"
 						delays.observe({dsn: dsn, status: status}, delay)
 					end
+
+					messages.increment(type: "delay")
+				elsif msg =~ %r{postfix/smtpd\[\d+\]: connect from }
+					connects.increment({})
+					active.send(:synchronize) { active.set({}, active.get({}) || 0 + 1) }
+					messages.increment(type: "connect")
+				elsif msg =~ %r{postfix/smtpd\[\d+\]: disconnect from }
+					active.send(:synchronize) do
+					  new = (active.get({}) || 0) - 1
+					  # If we start running mid-stream,
+					  # we might end up seeing more
+					  # disconnects than connections,
+					  # which would be confusing
+					  new = 0 if new < 0
+					  active.set({}, new)
+					end
+					messages.increment(type: "disconnect")
+				elsif msg =~ %r{postfix/smtpd\[\d+\]: [A-F0-9]+: client=}
+					incoming.increment(dsn: "2.0.0", status: "queued")
+					messages.increment(type: "queued")
+				elsif msg =~ %r{postfix/smtpd\[\d+\]: NOQUEUE: reject: RCPT from \S+: \d{3} (\d+\.\d+\.\d+) }
+					incoming.increment(dsn: $1, status: "rejected")
+					messages.increment(type: "noqueue")
+				else
+					messages.increment(type: "ignored")
 				end
 			rescue StandardError => ex
 				$stderr.puts "Error while receiving postfix logs: #{ex.message} (#{ex.class})"
 				$stderr.puts ex.backtrace.map { |l| "  #{l}" }.join("\n")
+				log_errors.increment(class: ex.class.to_s)
 				sleep 1
 			end
 		end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: postfix-exporter
 version: !ruby/object:Gem::Version
-  version: 2.0.3
+  version: 2.1.0
 platform: ruby
 authors:
 - Matt Palmer
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2017-07-05 00:00:00.000000000 Z
+date: 2018-01-30 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: prometheus-client
@@ -118,7 +118,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
       version: '0'
 requirements: []
 rubyforge_project: 
-rubygems_version: 2.5.2
+rubygems_version: 2.6.13
 signing_key: 
 specification_version: 4
 summary: Export Prometheus statistics for a Postfix server