scout_apm 2.6.6 → 4.0.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/.rubocop.yml +0 -4
- data/.travis.yml +0 -6
- data/CHANGELOG.markdown +30 -0
- data/Gemfile +1 -8
- data/lib/scout_apm.rb +21 -1
- data/lib/scout_apm/agent.rb +22 -0
- data/lib/scout_apm/agent_context.rb +14 -2
- data/lib/scout_apm/background_job_integrations/sidekiq.rb +2 -2
- data/lib/scout_apm/config.rb +17 -2
- data/lib/scout_apm/detailed_trace.rb +2 -1
- data/lib/scout_apm/error.rb +27 -0
- data/lib/scout_apm/error_service.rb +32 -0
- data/lib/scout_apm/error_service/error_buffer.rb +39 -0
- data/lib/scout_apm/error_service/error_record.rb +211 -0
- data/lib/scout_apm/error_service/ignored_exceptions.rb +66 -0
- data/lib/scout_apm/error_service/middleware.rb +32 -0
- data/lib/scout_apm/error_service/notifier.rb +33 -0
- data/lib/scout_apm/error_service/payload.rb +47 -0
- data/lib/scout_apm/error_service/periodic_work.rb +17 -0
- data/lib/scout_apm/error_service/railtie.rb +11 -0
- data/lib/scout_apm/error_service/sidekiq.rb +80 -0
- data/lib/scout_apm/extensions/transaction_callback_payload.rb +1 -1
- data/lib/scout_apm/instruments/action_controller_rails_3_rails4.rb +47 -26
- data/lib/scout_apm/instruments/action_view.rb +7 -2
- data/lib/scout_apm/instruments/active_record.rb +13 -28
- data/lib/scout_apm/middleware.rb +1 -1
- data/lib/scout_apm/reporter.rb +8 -3
- data/lib/scout_apm/serializers/payload_serializer_to_json.rb +28 -10
- data/lib/scout_apm/slow_policy/age_policy.rb +33 -0
- data/lib/scout_apm/slow_policy/percent_policy.rb +22 -0
- data/lib/scout_apm/slow_policy/percentile_policy.rb +24 -0
- data/lib/scout_apm/slow_policy/policy.rb +21 -0
- data/lib/scout_apm/slow_policy/speed_policy.rb +16 -0
- data/lib/scout_apm/slow_request_policy.rb +18 -77
- data/lib/scout_apm/utils/sql_sanitizer.rb +1 -0
- data/lib/scout_apm/utils/sql_sanitizer_regex.rb +1 -1
- data/lib/scout_apm/utils/sql_sanitizer_regex_1_8_7.rb +1 -0
- data/lib/scout_apm/version.rb +1 -1
- data/scout_apm.gemspec +6 -6
- data/test/unit/agent_context_test.rb +29 -0
- data/test/unit/error_service/error_buffer_test.rb +25 -0
- data/test/unit/error_service/ignored_exceptions_test.rb +49 -0
- data/test/unit/serializers/payload_serializer_test.rb +36 -0
- data/test/unit/slow_request_policy_test.rb +41 -13
- data/test/unit/sql_sanitizer_test.rb +7 -0
- metadata +25 -62
- data/lib/scout_apm/slow_job_policy.rb +0 -111
- data/test/unit/slow_job_policy_test.rb +0 -6
@@ -1,111 +0,0 @@
|
|
1
|
-
# Long running class that determines if, and in how much detail a potentially
|
2
|
-
# slow job should be recorded in
|
3
|
-
|
4
|
-
module ScoutApm
|
5
|
-
class SlowJobPolicy
|
6
|
-
CAPTURE_TYPES = [
|
7
|
-
CAPTURE_DETAIL = "capture_detail",
|
8
|
-
CAPTURE_NONE = "capture_none",
|
9
|
-
]
|
10
|
-
|
11
|
-
# Adjust speed points. See the function
|
12
|
-
POINT_MULTIPLIER_SPEED = 0.25
|
13
|
-
|
14
|
-
# For each minute we haven't seen an endpoint
|
15
|
-
POINT_MULTIPLIER_AGE = 0.25
|
16
|
-
|
17
|
-
# Outliers are worth up to "1000ms" of weight
|
18
|
-
POINT_MULTIPLIER_PERCENTILE = 1.0
|
19
|
-
|
20
|
-
# Points for an endpoint's who's throughput * response time is a large % of
|
21
|
-
# overall time spent processing requests
|
22
|
-
POINT_MULTIPLIER_PERCENT_TIME = 2.5
|
23
|
-
|
24
|
-
# A hash of Job Names to the last time we stored a slow trace for it.
|
25
|
-
#
|
26
|
-
# Defaults to a start time that is pretty close to application boot time.
|
27
|
-
# So the "age" of an endpoint we've never seen is the time the application
|
28
|
-
# has been running.
|
29
|
-
attr_reader :last_seen
|
30
|
-
|
31
|
-
# The AgentContext we're running in
|
32
|
-
attr_reader :context
|
33
|
-
|
34
|
-
def initialize(context)
|
35
|
-
@context = context
|
36
|
-
|
37
|
-
zero_time = Time.now
|
38
|
-
@last_seen = Hash.new { |h, k| h[k] = zero_time }
|
39
|
-
end
|
40
|
-
|
41
|
-
def stored!(request)
|
42
|
-
last_seen[request.unique_name] = Time.now
|
43
|
-
end
|
44
|
-
|
45
|
-
# Determine if this job trace should be fully analyzed by scoring it
|
46
|
-
# across several metrics, and then determining if that's good enough to
|
47
|
-
# make it into this minute's payload.
|
48
|
-
#
|
49
|
-
# Due to the combining nature of the agent & layaway file, there's no
|
50
|
-
# guarantee that a high scoring local champion will still be a winner when
|
51
|
-
# they go up to "regionals" and are compared against the other processes
|
52
|
-
# running on a node.
|
53
|
-
def score(request)
|
54
|
-
unique_name = request.unique_name
|
55
|
-
if unique_name == :unknown
|
56
|
-
return -1 # A negative score, should never be good enough to store.
|
57
|
-
end
|
58
|
-
|
59
|
-
total_time = request.root_layer.total_call_time
|
60
|
-
|
61
|
-
# How long has it been since we've seen this?
|
62
|
-
age = Time.now - last_seen[unique_name]
|
63
|
-
|
64
|
-
# What approximate percentile was this request?
|
65
|
-
percentile = context.request_histograms.approximate_quantile_of_value(unique_name, total_time)
|
66
|
-
|
67
|
-
percent_of_total_time = context.transaction_time_consumed.percent_of_total(unique_name)
|
68
|
-
|
69
|
-
return speed_points(total_time) + percentile_points(percentile) + age_points(age) + percent_time_points(percent_of_total_time)
|
70
|
-
end
|
71
|
-
|
72
|
-
private
|
73
|
-
|
74
|
-
# Time in seconds
|
75
|
-
# Logarithm keeps huge times from swamping the other metrics.
|
76
|
-
# 1+ is necessary to keep the log function in positive territory.
|
77
|
-
def speed_points(time)
|
78
|
-
Math.log(1 + time) * POINT_MULTIPLIER_SPEED
|
79
|
-
end
|
80
|
-
|
81
|
-
def percentile_points(percentile)
|
82
|
-
if percentile < 40
|
83
|
-
0.4 # Don't put much emphasis on capturing low percentiles.
|
84
|
-
elsif percentile < 60
|
85
|
-
1.4 # Highest here to get mean traces
|
86
|
-
elsif percentile < 90
|
87
|
-
0.7 # Between 60 & 90% is fine.
|
88
|
-
elsif percentile >= 90
|
89
|
-
1.4 # Highest here to get 90+%ile traces
|
90
|
-
else
|
91
|
-
# impossible.
|
92
|
-
percentile
|
93
|
-
end
|
94
|
-
end
|
95
|
-
|
96
|
-
def age_points(age)
|
97
|
-
age / 60.0 * POINT_MULTIPLIER_AGE
|
98
|
-
end
|
99
|
-
|
100
|
-
# Of the total time spent handling endpoints in this app, if this endpoint
|
101
|
-
# is a higher percent, it should get more points.
|
102
|
-
#
|
103
|
-
# A: 20 calls @ 100ms each => 2 seconds of total time
|
104
|
-
# B: 10 calls @ 100ms each => 1 second of total time
|
105
|
-
#
|
106
|
-
# Then A is 66% of the total call time
|
107
|
-
def percent_time_points(percent) # Scale 0.0 - 1.0
|
108
|
-
percent * POINT_MULTIPLIER_PERCENT_TIME
|
109
|
-
end
|
110
|
-
end
|
111
|
-
end
|