1 # -*- encoding: binary -*-
8 # Raindrops::Watcher is a stand-alone Rack application for watching
9 # any number of TCP and UNIX listeners (all of them by default).
11 # It depends on the {Aggregate RubyGem}[http://rubygems.org/gems/aggregate]
13 # In your Rack config.ru:
15 # run Raindrops::Watcher(options = {})
17 # It takes the following options hash:
19 # - :listeners - an array of listener names, (e.g. %w(0.0.0.0:80 /tmp/sock))
20 # - :delay - interval between stats updates in seconds (default: 1)
22 # Raindrops::Watcher is compatible any thread-safe/thread-aware Rack
23 # middleware. It does not work well with multi-process web servers
24 # but can be used to monitor them. It consumes minimal resources
25 # with the default :delay.
31 # Returns an HTML summary listing of all listen interfaces watched on
33 # === GET /active/$LISTENER.txt
35 # Returns a plain text summary + histogram with X-* HTTP headers for
38 # e.g.: curl http://raindrops-demo.bogomips.org/active/0.0.0.0%3A80.txt
40 # === GET /active/$LISTENER.html
42 # Returns an HTML summary + histogram with X-* HTTP headers for
45 # e.g.: curl http://raindrops-demo.bogomips.org/active/0.0.0.0%3A80.html
47 # === GET /queued/$LISTENER.txt
49 # Returns a plain text summary + histogram with X-* HTTP headers for
52 # e.g.: curl http://raindrops-demo.bogomips.org/queued/0.0.0.0%3A80.txt
54 # === GET /queued/$LISTENER.html
56 # Returns an HTML summary + histogram with X-* HTTP headers for
59 # e.g.: curl http://raindrops-demo.bogomips.org/queued/0.0.0.0%3A80.html
61 # === POST /reset/$LISTENER
63 # Resets the active and queued statistics for the given listener.
65 # === GET /tail/$LISTENER.txt?active_min=1&queued_min=1
67 # Streams chunked a response to the client.
68 # Interval is the preconfigured +:delay+ of the application (default 1 second)
70 # The response is plain text in the following format:
72 # ISO8601_TIMESTAMP LISTENER_NAME ACTIVE_COUNT QUEUED_COUNT LINEFEED
76 # - active_min - do not stream a line until this active count is reached
77 # - queued_min - do not stream a line until this queued count is reached
79 # == Response headers (mostly the same names as Raindrops::LastDataRecv)
81 # - X-Count - number of samples polled
82 # - X-Last-Reset - date since the last reset
84 # The following headers are only present if X-Count is greater than one.
86 # - X-Min - lowest number of connections recorded
87 # - X-Max - highest number of connections recorded
88 # - X-Mean - mean number of connections recorded
89 # - X-Std-Dev - standard deviation of connection count
90 # - X-Outliers-Low - number of low outliers (hopefully many for queued)
91 # - X-Outliers-High - number of high outliers (hopefully zero for queued)
92 # - X-Current - current number of connections
93 # - X-First-Peak-At - date of when X-Max was first reached
94 # - X-Last-Peak-At - date of when X-Max was last reached
98 # There is a server running this app at http://raindrops-demo.bogomips.org/
99 # The Raindrops::Middleware demo is also accessible at
100 # http://raindrops-demo.bogomips.org/_raindrops
102 # The demo server is only limited to 30 users, so be sure not to abuse it
103 # by using the /tail/ endpoint too much.
104 class Raindrops::Watcher
106 attr_reader :snapshot
108 include Raindrops::Linux
109 DOC_URL = "http://raindrops.bogomips.org/Raindrops/Watcher.html"
110 Peak = Struct.new(:first, :last)
112 def initialize(opts = {})
113 @tcp_listeners = @unix_listeners = nil
114 if l = opts[:listeners]
116 Array(l).each { |addr| (addr =~ %r{\A/} ? unix : tcp) << addr }
117 unless tcp.empty? && unix.empty?
119 @unix_listeners = unix
123 @agg_class = opts[:agg_class] || Aggregate
124 @start_time = Time.now.utc
125 @active = Hash.new { |h,k| h[k] = @agg_class.new }
126 @queued = Hash.new { |h,k| h[k] = @agg_class.new }
127 @resets = Hash.new { |h,k| h[k] = @start_time }
128 @peak_active = Hash.new { |h,k| h[k] = Peak.new(@start_time, @start_time) }
129 @peak_queued = Hash.new { |h,k| h[k] = Peak.new(@start_time, @start_time) }
130 @snapshot = [ @start_time, {} ]
131 @delay = opts[:delay] || 1
134 @cond = ConditionVariable.new
144 @start.synchronize { @thr ||= aggregator_thread(env["rack.logger"]) }
145 case env["REQUEST_METHOD"]
151 Rack::Response.new(["Method Not Allowed"], 405).finish
155 def aggregate!(agg_hash, peak_hash, addr, number, now)
157 if (max = agg.max) && number > 0 && number >= max
158 peak = peak_hash[addr]
159 peak.first = now if number > max
165 def aggregator_thread(logger) # :nodoc:
166 @socket = sock = Raindrops::InetDiagSocket.new
169 combined = tcp_listener_stats(@tcp_listeners, sock)
170 combined.merge!(unix_listener_stats(@unix_listeners))
173 combined.each do |addr,stats|
174 aggregate!(@active, @peak_active, addr, stats.active, now)
175 aggregate!(@queued, @peak_queued, addr, stats.queued, now)
177 @snapshot = [ now, combined ]
181 logger.error "#{e.class} #{e.inspect}"
182 end while sleep(@delay) && @socket
189 def non_existent_stats(time)
190 [ time, @start_time, @agg_class.new, 0, Peak.new(@start_time, @start_time) ]
193 def active_stats(addr) # :nodoc:
195 time, combined = @snapshot
196 stats = combined[addr] or return non_existent_stats(time)
197 tmp, peak = @active[addr], @peak_active[addr]
198 [ time, @resets[addr], tmp.dup, stats.active, peak ]
202 def queued_stats(addr) # :nodoc:
204 time, combined = @snapshot
205 stats = combined[addr] or return non_existent_stats(time)
206 tmp, peak = @queued[addr], @peak_queued[addr]
207 [ time, @resets[addr], tmp.dup, stats.queued, peak ]
224 def agg_to_hash(reset_at, agg, current, peak)
226 "X-Count" => agg.count.to_s,
227 "X-Min" => agg.min.to_s,
228 "X-Max" => agg.max.to_s,
229 "X-Mean" => agg.mean.to_s,
230 "X-Std-Dev" => std_dev(agg),
231 "X-Outliers-Low" => agg.outliers_low.to_s,
232 "X-Outliers-High" => agg.outliers_high.to_s,
233 "X-Last-Reset" => reset_at.httpdate,
234 "X-Current" => current.to_s,
235 "X-First-Peak-At" => peak.first.httpdate,
236 "X-Last-Peak-At" => peak.last.httpdate,
240 def histogram_txt(agg)
241 updated_at, reset_at, agg, current, peak = *agg
242 headers = agg_to_hash(reset_at, agg, current, peak)
244 headers["Content-Type"] = "text/plain"
245 headers["Expires"] = (updated_at + @delay).httpdate
246 headers["Content-Length"] = bytesize(body).to_s
247 [ 200, headers, [ body ] ]
250 def histogram_html(agg, addr)
251 updated_at, reset_at, agg, current, peak = *agg
252 headers = agg_to_hash(reset_at, agg, current, peak)
254 "<head><title>#{hostname} - #{escape_html addr}</title></head>" \
257 "<tr><td>#{k.gsub(/^X-/, '')}</td><td>#{v}</td></tr>"
258 }.join << "</table><pre>#{escape_html agg}</pre>" \
259 "<form action='/reset/#{escape addr}' method='post'>" \
260 "<input type='submit' name='x' value='reset' /></form>" \
262 headers["Content-Type"] = "text/html"
263 headers["Expires"] = (updated_at + @delay).httpdate
264 headers["Content-Length"] = bytesize(body).to_s
265 [ 200, headers, [ body ] ]
271 case env["PATH_INFO"]
274 when %r{\A/active/(.+)\.txt\z}
275 histogram_txt(active_stats(unescape($1)))
276 when %r{\A/active/(.+)\.html\z}
278 histogram_html(active_stats(addr), addr)
279 when %r{\A/queued/(.+)\.txt\z}
280 histogram_txt(queued_stats(unescape($1)))
281 when %r{\A/queued/(.+)\.html\z}
283 histogram_html(queued_stats(addr), addr)
284 when %r{\A/tail/(.+)\.txt\z}
285 tail(unescape($1), env)
298 Rack::Response.new(["Not Found"], 404).finish
302 case env["PATH_INFO"]
303 when %r{\A/reset/(.+)\z}
304 reset!(env, unescape($1))
310 def reset!(env, addr)
312 @active.include?(addr) or return not_found
315 @resets[addr] = Time.now.utc
318 req = Rack::Request.new(env)
319 res = Rack::Response.new
320 url = req.referer || "#{req.host_with_port}/"
322 res.content_type.replace "text/plain"
323 res.write "Redirecting to #{url}"
328 updated_at, all = snapshot
330 "Content-Type" => "text/html",
331 "Last-Modified" => updated_at.httpdate,
332 "Expires" => (updated_at + @delay).httpdate,
334 body = "<html><head>" \
335 "<title>#{hostname} - all interfaces</title>" \
336 "</head><body><h3>Updated at #{updated_at.iso8601}</h3>" \
338 "<th>address</th><th>active</th><th>queued</th><th>reset</th>" \
341 a[0] <=> b[0] # sort by addr
342 end.map do |addr,stats|
345 "<td><a href='/tail/#{e_addr}.txt' " \
346 "title='"tail" output in real time'" \
347 ">#{escape_html addr}</a></td>" \
348 "<td><a href='/active/#{e_addr}.html' " \
349 "title='show active connection stats'>#{stats.active}</a></td>" \
350 "<td><a href='/queued/#{e_addr}.html' " \
351 "title='show queued connection stats'>#{stats.queued}</a></td>" \
352 "<td><form action='/reset/#{e_addr}' method='post'>" \
353 "<input title='reset statistics' " \
354 "type='submit' name='x' value='x' /></form></td>" \
356 end.join << "</table>" \
358 "This is running the #{self.class}</a> service, see " \
359 "<a href='#{DOC_URL}'>#{DOC_URL}</a> " \
360 "for more information and options." \
363 headers["Content-Length"] = bytesize(body).to_s
364 [ 200, headers, [ body ] ]
368 Tailer.new(self, addr, env).finish
371 # This is the response body returned for "/tail/$ADDRESS.txt". This
372 # must use a multi-threaded Rack server with streaming response support.
373 # It is an internal class and not expected to be used directly
375 def initialize(rdmon, addr, env) # :nodoc:
378 q = Rack::Utils.parse_query env["QUERY_STRING"]
379 @active_min = q["active_min"].to_i
380 @queued_min = q["queued_min"].to_i
381 len = Rack::Utils.bytesize(addr)
383 @fmt = "%20s % #{len}s % 10u % 10u\n"
384 case env["HTTP_VERSION"]
394 "Content-Type" => "text/plain",
395 "Cache-Control" => "no-transform",
396 "Expires" => Time.at(0).httpdate,
398 headers["Transfer-Encoding"] = "chunked" if @chunk
399 [ 200, headers, self ]
402 # called by the Rack server
405 time, all = @rdmon.wait_snapshot
406 stats = all[@addr] or next
407 stats.queued >= @queued_min or next
408 stats.active >= @active_min or next
409 body = sprintf(@fmt, time.iso8601, @addr, stats.active, stats.queued)
410 body = "#{body.size.to_s(16)}\r\n#{body}\r\n" if @chunk
413 yield "0\r\n\r\n" if @chunk
417 # shuts down the background thread, only for tests