Class: Riemann::Tools::HttpCheck

Inherits:
Object
  • Object
show all
Includes:
Riemann::Tools
Defined in:
lib/riemann/tools/http_check.rb

Constant Summary collapse

REQUIRED_RUBY_VERSION = 
'2.7.0'

Constants included from Riemann::Tools

VERSION

Instance Attribute Summary

Attributes included from Riemann::Tools

#argv

Instance Method Summary collapse

Methods included from Riemann::Tools

#attributes, #endpoint_name, included, #options, #report, #riemann, #run

Constructor Details

#initialize ⇒ HttpCheck

Returns a new instance of HttpCheck.



38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
 
# File 'lib/riemann/tools/http_check.rb', line 38

def initialize
  super

  @resolve_queue = Queue.new
  @work_queue = Queue.new

  @resolvers = []
  @workers = []

  opts[:resolvers].times do
    @resolvers << Thread.new do
      loop do
        uri = @resolve_queue.pop
        Thread.exit unless uri

        host = uri.host

        addresses = Resolv::DNS.new.getaddresses(host)
        if addresses.empty?
          host = host[1...-1] if host[0] == '[' && host[-1] == ']'
          begin
            addresses << IPAddr.new(host)
          rescue IPAddr::InvalidAddressError
            # Ignore
          end
        end

        if opts[:ignored_asn].any?
          addresses.reject! do |address|
            address_belongs_to_ignored_asn?(address)
          end
        end

        next if addresses.empty?

        @work_queue.push([uri, addresses])
      end
    end
  end

  opts[:workers].times do
    @workers << Thread.new do
      loop do
        uri, addresses = @work_queue.pop
        Thread.exit unless uri

        test_uri_addresses(uri, addresses)
      end
    end
  end
end

Instance Method Details

#address_belongs_to_ignored_asn?(address) ⇒ Boolean

Returns:

  • (Boolean) 


90
91
92
93
94
95
96
97
98
99
100
101
102
103
 
# File 'lib/riemann/tools/http_check.rb', line 90

def address_belongs_to_ignored_asn?(address)
  begin
    require 'maxmind/geoip2'
  rescue LoadError
    raise StandardError, 'MaxMind::GeoIP2 is not available. Please install the maxmind-geoip2 gem for filtering by ASN.'
  end

  @reader ||= MaxMind::GeoIP2::Reader.new(database: opts[:geoip_asn_database])
  asn = @reader.asn(address.to_s)

  opts[:ignored_asn].include?(asn&.autonomous_system_number)
rescue MaxMind::GeoIP2::AddressNotFoundError
  false
end

#endpoint_report(http, uri, service) ⇒ Object 



329
330
331
332
333
334
335
336
 
# File 'lib/riemann/tools/http_check.rb', line 329

def endpoint_report(http, uri, service)
  {
    service: endpoint_service(http, uri, service),
    hostname: uri.host,
    address: http.ipaddr,
    port: uri.port,
  }
end

#endpoint_service(http, uri, service) ⇒ Object 



338
339
340
 
# File 'lib/riemann/tools/http_check.rb', line 338

def endpoint_service(http, uri, service)
  "get #{redact_uri(uri)} #{endpoint_name(IPAddr.new(http.ipaddr), http.port)} #{service}"
end

#get_request(uri) ⇒ Object 



198
199
200
201
202
203
204
 
# File 'lib/riemann/tools/http_check.rb', line 198

def get_request(uri)
  request = ::Net::HTTP::Get.new(uri, { 'user-agent' => opts[:user_agent] })

  request.basic_auth(uri.user, uri.password)

  request
end

#latency_state(name, latency) ⇒ Object 



314
315
316
317
318
319
320
321
322
323
324
325
326
327
 
# File 'lib/riemann/tools/http_check.rb', line 314

def latency_state(name, latency)
  critical_threshold = opts[:"#{name}_latency_critical"]
  warning_threshold = opts[:"#{name}_latency_warning"]

  return if critical_threshold.zero? || warning_threshold.zero?

  if latency.nil? || latency > critical_threshold
    'critical'
  elsif latency > warning_threshold
    'warning'
  else
    'ok'
  end
end

#redact_uri(uri) ⇒ Object 



346
347
348
349
350
 
# File 'lib/riemann/tools/http_check.rb', line 346

def redact_uri(uri)
  reported_uri = uri.dup
  reported_uri.password = '**redacted**' if reported_uri.password
  reported_uri
end

#redirect_uri(uri, location) ⇒ Object 



251
252
253
254
255
256
257
258
259
260
261
 
# File 'lib/riemann/tools/http_check.rb', line 251

def redirect_uri(uri, location)
  res = URI.parse(location)

  res.scheme   ||= uri.scheme
  res.host     ||= uri.host
  res.port     ||= uri.port
  res.user     ||= res.user
  res.password ||= res.password

  res
end

#report_http_endpoint_latency(http, uri, latency, start, stop) ⇒ Object 



285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
 
# File 'lib/riemann/tools/http_check.rb', line 285

def report_http_endpoint_latency(http, uri, latency, start, stop)
  if stop
    metric = stop - start
    report(
      {
        state: latency_state(latency, metric),
        metric: metric,
        description: format('%.3f ms', metric * 1000),
      }.merge(endpoint_report(http, uri, "#{latency} latency")),
    )
  else
    report(
      {
        state: latency_state(latency, nil),
        description: 'timeout',
      }.merge(endpoint_report(http, uri, "#{latency} latency")),
    )
  end
end

#report_http_endpoint_max_redirects(http, uri) ⇒ Object 



305
306
307
308
309
310
311
312
 
# File 'lib/riemann/tools/http_check.rb', line 305

def report_http_endpoint_max_redirects(http, uri)
  report(
    {
      state: 'critical',
      description: "Reached the limit of #{opts[:max_redirects]} redirects",
    }.merge(endpoint_report(http, uri, 'redirects')),
  )
end

#report_http_endpoint_response_code(http, uri, response) ⇒ Object 



269
270
271
272
273
274
275
276
277
278
279
 
# File 'lib/riemann/tools/http_check.rb', line 269

def report_http_endpoint_response_code(http, uri, response)
  return unless response

  report(
    {
      state: response_code_state(response.code),
      metric: response.code.to_i,
      description: "#{response.code} #{response.message}",
    }.merge(endpoint_report(http, uri, 'response code')),
  )
end

#response_code_state(code) ⇒ Object 



281
282
283
 
# File 'lib/riemann/tools/http_check.rb', line 281

def response_code_state(code)
  opts[:response].include?(code) ? 'ok' : 'critical'
end

#same_origin?(left, right) ⇒ Boolean

Returns:

  • (Boolean) 


263
264
265
266
267
 
# File 'lib/riemann/tools/http_check.rb', line 263

def same_origin?(left, right)
  left.scheme == right.scheme &&
    left.host == right.host &&
    left.port == right.port
end

#service(uri, service) ⇒ Object 



342
343
344
 
# File 'lib/riemann/tools/http_check.rb', line 342

def service(uri, service)
  "get #{redact_uri(uri)} #{service}"
end

#shutdown ⇒ Object

Under normal operation, we have a single instance of this class for the lifetime of the process. But when testing, we create a new instance for each test, each with its resolvers and worker threads. The test process may end-up with a lot of running threads, hitting the OS limit of max threads by process and being unable to create more thread:

ThreadError: can’t create Thread: Resource temporarily unavailable

To avoid this situation, we provide this method.



114
115
116
117
118
119
120
 
# File 'lib/riemann/tools/http_check.rb', line 114

def shutdown
  @resolve_queue.close
  @resolvers.map(&:join)

  @work_queue.close
  @workers.map(&:join)
end

#test_uri_address(uri, address, request, redirect_count: 0) ⇒ Object 



206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
 
# File 'lib/riemann/tools/http_check.rb', line 206

def test_uri_address(uri, address, request, redirect_count: 0)
  response = nil

  start = Time.now
  connected = nil
  done = nil

  http = nil
  begin
    Timeout.timeout(opts[:http_timeout]) do
      http = ::Net::HTTP.start(uri.host, uri.port, use_ssl: uri.scheme == 'https', verify_mode: OpenSSL::SSL::VERIFY_NONE, ipaddr: address)
      connected = Time.now
      response = http.request(request)
    end
  rescue Timeout::Error
    # Ignore
  else
    done = Time.now
  ensure
    http&.finish
  end

  report_http_endpoint_response_code(http, uri, response) if opts[:checks].include?('response-code')
  report_http_endpoint_latency(http, uri, 'connection', start, connected) if opts[:checks].include?('connection-latency')
  report_http_endpoint_latency(http, uri, 'response', start, done) if opts[:checks].include?('response-latency')

  if opts[:follow_redirects] && %w[301 302].include?(response.code)
    next_uri = redirect_uri(uri, response['Location'])

    if same_origin?(uri, next_uri)
      if redirect_count == opts[:max_redirects]
        report_http_endpoint_max_redirects(http, uri)
        return nil
      else
        response = test_uri_address(next_uri, address, get_request(next_uri), redirect_count: redirect_count + 1)
      end
    end
  end

  response
rescue StandardError
  # Ignore this address
  nil
end

#test_uri_addresses(uri, addresses) ⇒ Object 



153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
 
# File 'lib/riemann/tools/http_check.rb', line 153

def test_uri_addresses(uri, addresses)
  request = get_request(uri)

  responses = addresses.map do |address|
    test_uri_address(uri, address.to_s, request)
  end

  responses.compact!

  return unless opts[:checks].include?('consistency')

  raise StandardError, "Could not get any response from #{uri.host}" unless responses.any?

  uniq_code = responses.map(&:code).uniq
  uniq_body = responses.map(&:body).uniq

  issues = []
  issues << "#{uniq_code.count} different response code" unless uniq_code.one?
  issues << "#{uniq_body.count} different response body" unless uniq_body.one?

  if issues.none?
    state = 'ok'
    description = "consistent response on all #{responses.count} endpoints"
  else
    state = 'critical'
    description = "#{issues.join(' and ')} on #{responses.count} endpoints"
  end

  report(
    service: service(uri, 'consistency'),
    state: state,
    description: description,
    hostname: uri.host,
    port: uri.port,
  )
rescue StandardError => e
  report(
    service: service(uri, 'consistency'),
    state: 'critical',
    description: e.message,
    hostname: uri.host,
    port: uri.port,
  )
end

#tick ⇒ Object 



122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
 
# File 'lib/riemann/tools/http_check.rb', line 122

def tick
  report(
    service: 'riemann http-check resolvers utilization',
    metric: (opts[:resolvers].to_f - @resolve_queue.num_waiting) / opts[:resolvers],
    state: @resolve_queue.num_waiting.positive? ? 'ok' : 'critical',
    tags: %w[riemann],
  )
  report(
    service: 'riemann http-check resolvers saturation',
    metric: @resolve_queue.length,
    state: @resolve_queue.empty? ? 'ok' : 'critical',
    tags: %w[riemann],
  )
  report(
    service: 'riemann http-check workers utilization',
    metric: (opts[:workers].to_f - @work_queue.num_waiting) / opts[:workers],
    state: @work_queue.num_waiting.positive? ? 'ok' : 'critical',
    tags: %w[riemann],
  )
  report(
    service: 'riemann http-check workers saturation',
    metric: @work_queue.length,
    state: @work_queue.empty? ? 'ok' : 'critical',
    tags: %w[riemann],
  )

  opts[:uri].each do |uri|
    @resolve_queue.push(URI(uri))
  end
end