nerve_pharmeasy 0.7.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +7 -0
- data/.gitignore +22 -0
- data/.mailmap +2 -0
- data/.nerve.rc +2 -0
- data/.travis.yml +8 -0
- data/CONTRIBUTING.md +28 -0
- data/Gemfile +2 -0
- data/Gemfile.lock +75 -0
- data/LICENSE.txt +22 -0
- data/README.md +116 -0
- data/Rakefile +7 -0
- data/Vagrantfile +121 -0
- data/bin/nerve +16 -0
- data/example/nerve.conf.json +54 -0
- data/example/nerve_services/etcd_service1.json +19 -0
- data/example/nerve_services/zookeeper_service1.json +18 -0
- data/lib/nerve/configuration_manager.rb +106 -0
- data/lib/nerve/log.rb +24 -0
- data/lib/nerve/reporter/base.rb +61 -0
- data/lib/nerve/reporter/etcd.rb +73 -0
- data/lib/nerve/reporter/zookeeper.rb +101 -0
- data/lib/nerve/reporter.rb +18 -0
- data/lib/nerve/ring_buffer.rb +30 -0
- data/lib/nerve/service_watcher/base.rb +65 -0
- data/lib/nerve/service_watcher/http.rb +70 -0
- data/lib/nerve/service_watcher/rabbitmq.rb +68 -0
- data/lib/nerve/service_watcher/tcp.rb +56 -0
- data/lib/nerve/service_watcher.rb +152 -0
- data/lib/nerve/utils.rb +17 -0
- data/lib/nerve/version.rb +3 -0
- data/lib/nerve.rb +249 -0
- data/nerve.conf.json +23 -0
- data/nerve.gemspec +33 -0
- data/spec/.gitkeep +0 -0
- data/spec/configuration_manager_spec.rb +31 -0
- data/spec/example_services_spec.rb +42 -0
- data/spec/factories/check.rb +16 -0
- data/spec/factories/service.rb +26 -0
- data/spec/lib/nerve/reporter_etcd_spec.rb +18 -0
- data/spec/lib/nerve/reporter_spec.rb +86 -0
- data/spec/lib/nerve/reporter_zookeeper_spec.rb +32 -0
- data/spec/lib/nerve/service_watcher_spec.rb +89 -0
- data/spec/lib/nerve_spec.rb +186 -0
- data/spec/spec_helper.rb +33 -0
- metadata +216 -0
@@ -0,0 +1,61 @@
|
|
1
|
+
|
2
|
+
class Nerve::Reporter
|
3
|
+
class Base
|
4
|
+
include Nerve::Utils
|
5
|
+
include Nerve::Logging
|
6
|
+
|
7
|
+
def initialize(opts)
|
8
|
+
end
|
9
|
+
|
10
|
+
def start
|
11
|
+
end
|
12
|
+
|
13
|
+
def stop
|
14
|
+
end
|
15
|
+
|
16
|
+
def report_up
|
17
|
+
end
|
18
|
+
|
19
|
+
def report_down
|
20
|
+
end
|
21
|
+
|
22
|
+
def ping?
|
23
|
+
end
|
24
|
+
|
25
|
+
def get_service_data(service)
|
26
|
+
%w{instance_id host port}.each do |required|
|
27
|
+
raise ArgumentError, "missing required argument #{required} for new service watcher" unless service[required]
|
28
|
+
end
|
29
|
+
d = {
|
30
|
+
'host' => service['host'],
|
31
|
+
'port' => service['port'],
|
32
|
+
'name' => service['instance_id']
|
33
|
+
}
|
34
|
+
|
35
|
+
# Weight is optional, but it should be well formed if supplied
|
36
|
+
if service.has_key?('weight')
|
37
|
+
if service['weight'].to_i >= 0 and "#{service['weight']}".match /^\d+$/
|
38
|
+
d['weight'] = service['weight'].to_i
|
39
|
+
else
|
40
|
+
raise ArgumentError, "invalid 'weight' argument in service data: #{service.inspect}"
|
41
|
+
end
|
42
|
+
end
|
43
|
+
|
44
|
+
if service.has_key?('haproxy_server_options')
|
45
|
+
d['haproxy_server_options'] = service['haproxy_server_options']
|
46
|
+
end
|
47
|
+
|
48
|
+
if service.has_key?('labels')
|
49
|
+
d['labels'] = service['labels']
|
50
|
+
end
|
51
|
+
d
|
52
|
+
end
|
53
|
+
|
54
|
+
protected
|
55
|
+
def parse_data(data)
|
56
|
+
return data if data.class == String
|
57
|
+
return data.to_json
|
58
|
+
end
|
59
|
+
end
|
60
|
+
end
|
61
|
+
|
@@ -0,0 +1,73 @@
|
|
1
|
+
require 'nerve/reporter/base'
|
2
|
+
require 'etcd'
|
3
|
+
|
4
|
+
class Nerve::Reporter
|
5
|
+
class Etcd < Base
|
6
|
+
def initialize(service)
|
7
|
+
raise ArgumentError, "missing required argument etcd_host for new service watcher" unless service['etcd_host']
|
8
|
+
@host = service['etcd_host']
|
9
|
+
@port = service['etcd_port'] || 4003
|
10
|
+
path = service['etcd_path'] || '/'
|
11
|
+
@path = path.split('/').push(service['instance_id']).join('/')
|
12
|
+
@data = parse_data(get_service_data(service))
|
13
|
+
@key = nil
|
14
|
+
@ttl = (service['check_interval'] || 0.5) * 5
|
15
|
+
@ttl = @ttl.ceil
|
16
|
+
end
|
17
|
+
|
18
|
+
def start()
|
19
|
+
log.info "nerve: connecting to etcd at #{@host}:#{@port}"
|
20
|
+
@etcd = ::Etcd.client(:host => @host, :port => @port)
|
21
|
+
log.info "nerve: successfully created etcd connection to #{@host}:#{@port}"
|
22
|
+
end
|
23
|
+
|
24
|
+
def stop()
|
25
|
+
report_down
|
26
|
+
@etcd = nil
|
27
|
+
end
|
28
|
+
|
29
|
+
def report_up()
|
30
|
+
etcd_save
|
31
|
+
end
|
32
|
+
|
33
|
+
def report_down
|
34
|
+
etcd_delete
|
35
|
+
end
|
36
|
+
|
37
|
+
def ping?
|
38
|
+
# we get a ping every check_interval.
|
39
|
+
if @key
|
40
|
+
# we have made a key: save it to prevent the TTL from expiring.
|
41
|
+
etcd_save
|
42
|
+
else
|
43
|
+
# we haven't created a key, so just frob the etcd API to assure that
|
44
|
+
# it's alive.
|
45
|
+
@etcd.leader
|
46
|
+
end
|
47
|
+
end
|
48
|
+
|
49
|
+
private
|
50
|
+
|
51
|
+
def etcd_delete
|
52
|
+
return unless @etcd and @key
|
53
|
+
begin
|
54
|
+
@etcd.delete(@key)
|
55
|
+
rescue ::Etcd::NotFile
|
56
|
+
rescue Errno::ECONNREFUSED
|
57
|
+
end
|
58
|
+
end
|
59
|
+
|
60
|
+
def etcd_create
|
61
|
+
# we use create_in_order to create a unique key under our path,
|
62
|
+
# permitting multiple registrations from the same instance_id.
|
63
|
+
@key = @etcd.create_in_order(@path, :value => @data, :ttl => @ttl).key
|
64
|
+
log.info "registered etcd key #{@key} with value #{@data}, TTL #{@ttl}"
|
65
|
+
end
|
66
|
+
|
67
|
+
def etcd_save
|
68
|
+
return etcd_create unless @key
|
69
|
+
@etcd.set(@key, :value => @data, :ttl => @ttl)
|
70
|
+
end
|
71
|
+
end
|
72
|
+
end
|
73
|
+
|
@@ -0,0 +1,101 @@
|
|
1
|
+
require 'nerve/reporter/base'
|
2
|
+
require 'thread'
|
3
|
+
require 'zk'
|
4
|
+
|
5
|
+
|
6
|
+
class Nerve::Reporter
|
7
|
+
class Zookeeper < Base
|
8
|
+
@@zk_pool = {}
|
9
|
+
@@zk_pool_count = {}
|
10
|
+
@@zk_pool_lock = Mutex.new
|
11
|
+
|
12
|
+
def initialize(service)
|
13
|
+
%w{zk_hosts zk_path}.each do |required|
|
14
|
+
raise ArgumentError, "missing required argument #{required} for new service watcher" unless service[required]
|
15
|
+
end
|
16
|
+
# Since we pool we get one connection per zookeeper cluster
|
17
|
+
@zk_connection_string = service['zk_hosts'].sort.join(',')
|
18
|
+
@data = parse_data(get_service_data(service))
|
19
|
+
|
20
|
+
@zk_path = service['zk_path']
|
21
|
+
@key_prefix = @zk_path + "/#{service['instance_id']}_"
|
22
|
+
@full_key = nil
|
23
|
+
end
|
24
|
+
|
25
|
+
def start()
|
26
|
+
log.info "nerve: waiting to connect to zookeeper to #{@zk_connection_string}"
|
27
|
+
# Ensure that all Zookeeper reporters re-use a single zookeeper
|
28
|
+
# connection to any given set of zk hosts.
|
29
|
+
@@zk_pool_lock.synchronize {
|
30
|
+
unless @@zk_pool.has_key?(@zk_connection_string)
|
31
|
+
log.info "nerve: creating pooled connection to #{@zk_connection_string}"
|
32
|
+
@@zk_pool[@zk_connection_string] = ZK.new(@zk_connection_string, :timeout => 5)
|
33
|
+
@@zk_pool_count[@zk_connection_string] = 1
|
34
|
+
log.info "nerve: successfully created zk connection to #{@zk_connection_string}"
|
35
|
+
else
|
36
|
+
@@zk_pool_count[@zk_connection_string] += 1
|
37
|
+
log.info "nerve: re-using existing zookeeper connection to #{@zk_connection_string}"
|
38
|
+
end
|
39
|
+
@zk = @@zk_pool[@zk_connection_string]
|
40
|
+
log.info "nerve: retrieved zk connection to #{@zk_connection_string}"
|
41
|
+
}
|
42
|
+
end
|
43
|
+
|
44
|
+
def stop()
|
45
|
+
log.info "nerve: removing zk node at #{@full_key}" if @full_key
|
46
|
+
begin
|
47
|
+
report_down
|
48
|
+
ensure
|
49
|
+
@@zk_pool_lock.synchronize {
|
50
|
+
@@zk_pool_count[@zk_connection_string] -= 1
|
51
|
+
# Last thread to use the connection closes it
|
52
|
+
if @@zk_pool_count[@zk_connection_string] == 0
|
53
|
+
log.info "nerve: closing zk connection to #{@zk_connection_string}"
|
54
|
+
begin
|
55
|
+
@zk.close!
|
56
|
+
ensure
|
57
|
+
@@zk_pool.delete(@zk_connection_string)
|
58
|
+
end
|
59
|
+
end
|
60
|
+
}
|
61
|
+
end
|
62
|
+
end
|
63
|
+
|
64
|
+
def report_up()
|
65
|
+
zk_save
|
66
|
+
end
|
67
|
+
|
68
|
+
def report_down
|
69
|
+
zk_delete
|
70
|
+
end
|
71
|
+
|
72
|
+
def ping?
|
73
|
+
return @zk.connected? && @zk.exists?(@full_key || '/')
|
74
|
+
end
|
75
|
+
|
76
|
+
private
|
77
|
+
|
78
|
+
def zk_delete
|
79
|
+
if @full_key
|
80
|
+
@zk.delete(@full_key, :ignore => :no_node)
|
81
|
+
@full_key = nil
|
82
|
+
end
|
83
|
+
end
|
84
|
+
|
85
|
+
def zk_create
|
86
|
+
@zk.mkdir_p(@zk_path)
|
87
|
+
@full_key = @zk.create(@key_prefix, :data => @data, :mode => :ephemeral_sequential)
|
88
|
+
end
|
89
|
+
|
90
|
+
def zk_save
|
91
|
+
return zk_create unless @full_key
|
92
|
+
|
93
|
+
begin
|
94
|
+
@zk.set(@full_key, @data)
|
95
|
+
rescue ZK::Exceptions::NoNode
|
96
|
+
zk_create
|
97
|
+
end
|
98
|
+
end
|
99
|
+
end
|
100
|
+
end
|
101
|
+
|
@@ -0,0 +1,18 @@
|
|
1
|
+
require 'nerve/utils'
|
2
|
+
require 'nerve/log'
|
3
|
+
require 'nerve/reporter/base'
|
4
|
+
|
5
|
+
module Nerve
|
6
|
+
class Reporter
|
7
|
+
def self.new_from_service(service)
|
8
|
+
type = service['reporter_type'] || 'zookeeper'
|
9
|
+
reporter = begin
|
10
|
+
require "nerve/reporter/#{type.downcase}"
|
11
|
+
self.const_get(type.downcase.capitalize)
|
12
|
+
rescue Exception => e
|
13
|
+
raise ArgumentError, "specified a reporter_type of #{type}, which could not be found: #{e}"
|
14
|
+
end
|
15
|
+
reporter.new(service)
|
16
|
+
end
|
17
|
+
end
|
18
|
+
end
|
@@ -0,0 +1,30 @@
|
|
1
|
+
module Nerve
|
2
|
+
class RingBuffer < Array
|
3
|
+
alias_method :array_push, :push
|
4
|
+
alias_method :array_element, :[]
|
5
|
+
|
6
|
+
def initialize( size )
|
7
|
+
@ring_size = size.to_i
|
8
|
+
super( @ring_size )
|
9
|
+
end
|
10
|
+
|
11
|
+
def average
|
12
|
+
self.inject(0.0) { |sum, el| sum + el } / self.size
|
13
|
+
end
|
14
|
+
|
15
|
+
def push( element )
|
16
|
+
if length == @ring_size
|
17
|
+
shift # loose element
|
18
|
+
end
|
19
|
+
array_push element
|
20
|
+
end
|
21
|
+
|
22
|
+
# Access elements in the RingBuffer
|
23
|
+
#
|
24
|
+
# offset will be typically negative!
|
25
|
+
#
|
26
|
+
def []( offset = 0 )
|
27
|
+
return self.array_element( - 1 + offset )
|
28
|
+
end
|
29
|
+
end
|
30
|
+
end
|
@@ -0,0 +1,65 @@
|
|
1
|
+
require 'nerve/ring_buffer'
|
2
|
+
|
3
|
+
module Nerve
|
4
|
+
module ServiceCheck
|
5
|
+
class BaseServiceCheck
|
6
|
+
include Utils
|
7
|
+
include Logging
|
8
|
+
|
9
|
+
def initialize(opts={})
|
10
|
+
@timeout = opts['timeout'] ? opts['timeout'].to_f : 0.1
|
11
|
+
@rise = opts['rise'] ? opts['rise'].to_i : 1
|
12
|
+
@fall = opts['fall'] ? opts['fall'].to_i : 1
|
13
|
+
@name = opts['name'] ? opts['name'] : "undefined"
|
14
|
+
|
15
|
+
@check_buffer = RingBuffer.new([@rise, @fall].max)
|
16
|
+
@last_result = nil
|
17
|
+
end
|
18
|
+
|
19
|
+
def up?
|
20
|
+
# do the check
|
21
|
+
check_result = !!catch_errors do
|
22
|
+
check
|
23
|
+
end
|
24
|
+
|
25
|
+
# this is the first check -- initialize buffer
|
26
|
+
if @last_result == nil
|
27
|
+
@last_result = check_result
|
28
|
+
@check_buffer.size.times {@check_buffer.push check_result}
|
29
|
+
log.info "nerve: service check #{@name} initial check returned #{check_result}"
|
30
|
+
end
|
31
|
+
|
32
|
+
log.debug "nerve: service check #{@name} returned #{check_result}"
|
33
|
+
@check_buffer.push(check_result)
|
34
|
+
|
35
|
+
# we've failed if the last @fall times are false
|
36
|
+
unless @check_buffer.last(@fall).reduce(:|)
|
37
|
+
log.info "nerve: service check #{@name} transitions to down after #{@fall} failures" if @last_result
|
38
|
+
@last_result = false
|
39
|
+
end
|
40
|
+
|
41
|
+
# we've succeeded if the last @rise times is true
|
42
|
+
if @check_buffer.last(@rise).reduce(:&)
|
43
|
+
log.info "nerve: service check #{@name} transitions to up after #{@rise} successes" unless @last_result
|
44
|
+
@last_result = true
|
45
|
+
end
|
46
|
+
|
47
|
+
# otherwise return the last result
|
48
|
+
return @last_result
|
49
|
+
end
|
50
|
+
|
51
|
+
def catch_errors(&block)
|
52
|
+
begin
|
53
|
+
return yield
|
54
|
+
rescue Object => error
|
55
|
+
log.info "nerve: service check #{@name} got error #{error.inspect}"
|
56
|
+
return false
|
57
|
+
end
|
58
|
+
end
|
59
|
+
end
|
60
|
+
|
61
|
+
CHECKS ||= {}
|
62
|
+
CHECKS['base'] = BaseServiceCheck
|
63
|
+
end
|
64
|
+
end
|
65
|
+
|
@@ -0,0 +1,70 @@
|
|
1
|
+
require 'nerve/service_watcher/base'
|
2
|
+
|
3
|
+
module Nerve
|
4
|
+
module ServiceCheck
|
5
|
+
require 'net/http'
|
6
|
+
|
7
|
+
class HttpServiceCheck < BaseServiceCheck
|
8
|
+
def initialize(opts={})
|
9
|
+
super
|
10
|
+
|
11
|
+
%w{port uri}.each do |required|
|
12
|
+
raise ArgumentError, "missing required argument #{required} in http check" unless
|
13
|
+
opts[required]
|
14
|
+
instance_variable_set("@#{required}",opts[required])
|
15
|
+
end
|
16
|
+
|
17
|
+
@host = opts['host'] || '127.0.0.1'
|
18
|
+
@ssl = opts['ssl'] || false
|
19
|
+
|
20
|
+
@read_timeout = opts['read_timeout'] || @timeout
|
21
|
+
@open_timeout = opts['open_timeout'] || 0.2
|
22
|
+
@ssl_timeout = opts['ssl_timeout'] || 0.2
|
23
|
+
|
24
|
+
@headers = opts['headers'] || {}
|
25
|
+
|
26
|
+
@expect = opts['expect']
|
27
|
+
|
28
|
+
@name = "http-#{@host}:#{@port}#{@uri}"
|
29
|
+
end
|
30
|
+
|
31
|
+
def check
|
32
|
+
log.debug "running health check #{@name}"
|
33
|
+
|
34
|
+
connection = get_connection
|
35
|
+
response = connection.get(@uri, @headers)
|
36
|
+
code = response.code.to_i
|
37
|
+
body = response.body
|
38
|
+
|
39
|
+
# Any 2xx or 3xx code should be considered healthy. This is standard
|
40
|
+
# practice in HAProxy, nginx, etc ...
|
41
|
+
if code >= 200 and code < 400 and (@expect == nil || body.include?(@expect))
|
42
|
+
log.debug "nerve: check #{@name} got response code #{code} with body \"#{body}\""
|
43
|
+
return true
|
44
|
+
else
|
45
|
+
log.warn "nerve: check #{@name} got response code #{code} with body \"#{body}\""
|
46
|
+
return false
|
47
|
+
end
|
48
|
+
end
|
49
|
+
|
50
|
+
private
|
51
|
+
def get_connection
|
52
|
+
con = Net::HTTP.new(@host, @port)
|
53
|
+
con.read_timeout = @read_timeout
|
54
|
+
con.open_timeout = @open_timeout
|
55
|
+
|
56
|
+
if @ssl
|
57
|
+
con.use_ssl = true
|
58
|
+
con.ssl_timeout = @ssl_timeout
|
59
|
+
con.verify_mode = OpenSSL::SSL::VERIFY_NONE
|
60
|
+
end
|
61
|
+
|
62
|
+
return con
|
63
|
+
end
|
64
|
+
|
65
|
+
end
|
66
|
+
|
67
|
+
CHECKS ||= {}
|
68
|
+
CHECKS['http'] = HttpServiceCheck
|
69
|
+
end
|
70
|
+
end
|
@@ -0,0 +1,68 @@
|
|
1
|
+
require 'nerve/service_watcher/base'
|
2
|
+
require 'bunny'
|
3
|
+
|
4
|
+
module Nerve
|
5
|
+
module ServiceCheck
|
6
|
+
class RabbitMQServiceCheck < BaseServiceCheck
|
7
|
+
require 'socket'
|
8
|
+
include Socket::Constants
|
9
|
+
|
10
|
+
def initialize(opts={})
|
11
|
+
super
|
12
|
+
|
13
|
+
raise ArgumentError, "missing required argument 'port' in rabbitmq check" unless opts['port']
|
14
|
+
|
15
|
+
@port = opts['port']
|
16
|
+
@host = opts['host'] || '127.0.0.1'
|
17
|
+
@user = opts['username'] || 'guest'
|
18
|
+
@pass = opts['password'] || 'guest'
|
19
|
+
end
|
20
|
+
|
21
|
+
def check
|
22
|
+
# the idea for this check was taken from the one in rabbitmq management
|
23
|
+
# -- the aliveness_test:
|
24
|
+
# https://github.com/rabbitmq/rabbitmq-management/blob/9a8e3d1ab5144e3f6a1cb9a4639eb738713b926d/src/rabbit_mgmt_wm_aliveness_test.erl
|
25
|
+
log.debug "nerve: running rabbitmq health check #{@name}"
|
26
|
+
|
27
|
+
conn = Bunny.new(
|
28
|
+
:host => @host,
|
29
|
+
:port => @port,
|
30
|
+
:user => @user,
|
31
|
+
:pass => @pass,
|
32
|
+
:log_file => STDERR,
|
33
|
+
:continuation_timeout => @timeout,
|
34
|
+
:automatically_recover => false,
|
35
|
+
:heartbeat => false,
|
36
|
+
:threaded => false
|
37
|
+
)
|
38
|
+
|
39
|
+
begin
|
40
|
+
conn.start
|
41
|
+
ch = conn.create_channel
|
42
|
+
|
43
|
+
# create a queue, publish to it
|
44
|
+
log.debug "nerve: publishing to rabbitmq"
|
45
|
+
ch.queue('nerve')
|
46
|
+
ch.basic_publish('nerve test message', '', 'nerve', :mandatory => true, :expiration => 2 * 1000)
|
47
|
+
|
48
|
+
# read and ack the message
|
49
|
+
log.debug "nerve: consuming from rabbitmq"
|
50
|
+
delivery_info, properties, payload = ch.basic_get('nerve', :ack => true)
|
51
|
+
|
52
|
+
if payload
|
53
|
+
ch.acknowledge(delivery_info.delivery_tag)
|
54
|
+
return true
|
55
|
+
else
|
56
|
+
log.debug "nerve: rabbitmq consumption returned no payload"
|
57
|
+
return false
|
58
|
+
end
|
59
|
+
ensure
|
60
|
+
conn.close
|
61
|
+
end
|
62
|
+
end
|
63
|
+
end
|
64
|
+
|
65
|
+
CHECKS ||= {}
|
66
|
+
CHECKS['rabbitmq'] = RabbitMQServiceCheck
|
67
|
+
end
|
68
|
+
end
|
@@ -0,0 +1,56 @@
|
|
1
|
+
require 'nerve/service_watcher/base'
|
2
|
+
|
3
|
+
module Nerve
|
4
|
+
module ServiceCheck
|
5
|
+
class TcpServiceCheck < BaseServiceCheck
|
6
|
+
require 'socket'
|
7
|
+
include Socket::Constants
|
8
|
+
|
9
|
+
def initialize(opts={})
|
10
|
+
super
|
11
|
+
|
12
|
+
raise ArgumentError, "missing required argument 'port' in tcp check" unless opts['port']
|
13
|
+
|
14
|
+
@port = opts['port']
|
15
|
+
@host = opts['host'] || '127.0.0.1'
|
16
|
+
|
17
|
+
@address = Socket.sockaddr_in(@port, @host)
|
18
|
+
end
|
19
|
+
|
20
|
+
def check
|
21
|
+
log.debug "nerve: running TCP health check #{@name}"
|
22
|
+
|
23
|
+
# create a TCP socket
|
24
|
+
socket = Socket.new(AF_INET, SOCK_STREAM, 0)
|
25
|
+
|
26
|
+
begin
|
27
|
+
# open a non-blocking connection
|
28
|
+
socket.connect_nonblock(@address)
|
29
|
+
rescue Errno::EINPROGRESS
|
30
|
+
# opening a non-blocking socket will usually raise
|
31
|
+
# this exception. it's just connect returning immediately,
|
32
|
+
# so it's not really an exception, but ruby makes it into
|
33
|
+
# one. if we got here, we are now free to wait until the timeout
|
34
|
+
# expires for the socket to be writeable
|
35
|
+
IO.select(nil, [socket], nil, @timeout)
|
36
|
+
|
37
|
+
# we should be connected now; allow any other exception through
|
38
|
+
begin
|
39
|
+
socket.connect_nonblock(@address)
|
40
|
+
rescue Errno::EISCONN
|
41
|
+
return true
|
42
|
+
end
|
43
|
+
else
|
44
|
+
# we managed to connect REALLY REALLY FAST
|
45
|
+
log.debug "nerve: connected to non-blocking socket without an exception"
|
46
|
+
return true
|
47
|
+
ensure
|
48
|
+
socket.close
|
49
|
+
end
|
50
|
+
end
|
51
|
+
end
|
52
|
+
|
53
|
+
CHECKS ||= {}
|
54
|
+
CHECKS['tcp'] = TcpServiceCheck
|
55
|
+
end
|
56
|
+
end
|
@@ -0,0 +1,152 @@
|
|
1
|
+
require 'nerve/service_watcher/tcp'
|
2
|
+
require 'nerve/service_watcher/http'
|
3
|
+
require 'nerve/service_watcher/rabbitmq'
|
4
|
+
|
5
|
+
module Nerve
|
6
|
+
class ServiceWatcher
|
7
|
+
include Utils
|
8
|
+
include Logging
|
9
|
+
|
10
|
+
attr_reader :was_up
|
11
|
+
|
12
|
+
def initialize(service={})
|
13
|
+
log.debug "nerve: creating service watcher object"
|
14
|
+
|
15
|
+
# check that we have all of the required arguments
|
16
|
+
%w{name instance_id host port}.each do |required|
|
17
|
+
raise ArgumentError, "missing required argument #{required} for new service watcher" unless service[required]
|
18
|
+
end
|
19
|
+
|
20
|
+
@name = service['name']
|
21
|
+
|
22
|
+
# configure the reporter, which we use for reporting status to the registry
|
23
|
+
@reporter = Reporter.new_from_service(service)
|
24
|
+
|
25
|
+
# instantiate the checks for this service
|
26
|
+
@service_checks = []
|
27
|
+
service['checks'] ||= []
|
28
|
+
service['checks'].each do |check|
|
29
|
+
# checks inherit attributes from the service overall
|
30
|
+
check['host'] ||= service['host']
|
31
|
+
check['port'] ||= service['port']
|
32
|
+
|
33
|
+
# generate a nice readable name for each check
|
34
|
+
check['name'] ||= "#{@name} #{check['type']}-#{check['host']}:#{check['port']}"
|
35
|
+
|
36
|
+
# make sure a type is set
|
37
|
+
check['type'] ||= "undefined"
|
38
|
+
|
39
|
+
# require a 3rd-party module if necessary for external checkers
|
40
|
+
unless ServiceCheck::CHECKS[check['type']]
|
41
|
+
m = check['module'] ? check['module'] : "nerve-watcher-#{check['type']}"
|
42
|
+
require m
|
43
|
+
end
|
44
|
+
|
45
|
+
# instantiate the check object
|
46
|
+
service_check_class = ServiceCheck::CHECKS[check['type']]
|
47
|
+
if service_check_class.nil?
|
48
|
+
raise ArgumentError,
|
49
|
+
"invalid service check type #{check['type']}; valid types: #{ServiceCheck::CHECKS.keys.join(',')}"
|
50
|
+
end
|
51
|
+
|
52
|
+
# save the check object
|
53
|
+
@service_checks << service_check_class.new(check)
|
54
|
+
end
|
55
|
+
|
56
|
+
# how often do we initiate service checks?
|
57
|
+
@check_interval = service['check_interval'] || 0.5
|
58
|
+
|
59
|
+
# force an initial report on startup
|
60
|
+
@was_up = nil
|
61
|
+
|
62
|
+
# when this watcher is started it will store the
|
63
|
+
# thread here
|
64
|
+
@run_thread = nil
|
65
|
+
@should_finish = false
|
66
|
+
|
67
|
+
log.debug "nerve: created service watcher for #{@name} with #{@service_checks.size} checks"
|
68
|
+
end
|
69
|
+
|
70
|
+
def start()
|
71
|
+
unless @run_thread
|
72
|
+
@run_thread = Thread.new { self.run() }
|
73
|
+
else
|
74
|
+
log.error "nerve: tried to double start a watcher"
|
75
|
+
end
|
76
|
+
end
|
77
|
+
|
78
|
+
def stop()
|
79
|
+
log.info "nerve: stopping service watch #{@name}"
|
80
|
+
@should_finish = true
|
81
|
+
return true if @run_thread.nil?
|
82
|
+
|
83
|
+
unclean_shutdown = @run_thread.join(10).nil?
|
84
|
+
if unclean_shutdown
|
85
|
+
log.error "nerve: unclean shutdown of #{@name}, killing thread"
|
86
|
+
Thread.kill(@run_thread)
|
87
|
+
end
|
88
|
+
@run_thread = nil
|
89
|
+
!unclean_shutdown
|
90
|
+
end
|
91
|
+
|
92
|
+
def alive?()
|
93
|
+
!@run_thread.nil? && @run_thread.alive?
|
94
|
+
end
|
95
|
+
|
96
|
+
def run()
|
97
|
+
log.info "nerve: starting service watch #{@name}"
|
98
|
+
@reporter.start()
|
99
|
+
|
100
|
+
until watcher_should_exit?
|
101
|
+
check_and_report
|
102
|
+
|
103
|
+
# wait to run more checks but make sure to exit if $EXIT
|
104
|
+
# we avoid sleeping for the entire check interval at once
|
105
|
+
# so that nerve can exit promptly if required
|
106
|
+
responsive_sleep (@check_interval) { watcher_should_exit? }
|
107
|
+
end
|
108
|
+
rescue StandardError => e
|
109
|
+
log.error "nerve: error in service watcher #{@name}: #{e.inspect}"
|
110
|
+
raise e
|
111
|
+
ensure
|
112
|
+
log.info "nerve: stopping reporter for #{@name}"
|
113
|
+
@reporter.stop
|
114
|
+
end
|
115
|
+
|
116
|
+
def check_and_report
|
117
|
+
if !@reporter.ping?
|
118
|
+
# If the reporter can't ping, then we do not know the status
|
119
|
+
# and must force a new report.
|
120
|
+
@was_up = nil
|
121
|
+
end
|
122
|
+
|
123
|
+
# what is the status of the service?
|
124
|
+
is_up = check?
|
125
|
+
log.debug "nerve: current service status for #{@name} is #{is_up.inspect}"
|
126
|
+
|
127
|
+
if is_up != @was_up
|
128
|
+
if is_up
|
129
|
+
@reporter.report_up
|
130
|
+
log.info "nerve: service #{@name} is now up"
|
131
|
+
else
|
132
|
+
@reporter.report_down
|
133
|
+
log.warn "nerve: service #{@name} is now down"
|
134
|
+
end
|
135
|
+
@was_up = is_up
|
136
|
+
end
|
137
|
+
end
|
138
|
+
|
139
|
+
def check?
|
140
|
+
@service_checks.each do |check|
|
141
|
+
return false unless check.up?
|
142
|
+
end
|
143
|
+
return true
|
144
|
+
end
|
145
|
+
|
146
|
+
private
|
147
|
+
def watcher_should_exit?
|
148
|
+
$EXIT || @should_finish
|
149
|
+
end
|
150
|
+
|
151
|
+
end
|
152
|
+
end
|