Tweet
Tweet

Slide 1

Slide 1 text

Hynek Schlawack Get Instrumented How Prometheus Can Unify Your Metrics

Slide 2

Slide 2 text

Goals

Slide 3

Slide 3 text

Goals

Slide 4

Slide 4 text

Goals

Slide 5

Slide 5 text

Goals

Slide 6

Slide 6 text

Goals

Slide 7

Slide 7 text

Service Level

Slide 8

Slide 8 text

Service Level Indicator

Slide 9

Slide 9 text

Service Level Indicator Objective

Slide 10

Slide 10 text

Service Level Indicator Objective (Agreement)

Slide 11

Slide 11 text

Metrics

Slide 12

Slide 12 text

Metrics avg latency 0.3 0.5 0.8 1.1 2.6

Slide 13

Slide 13 text

Metrics 12:00 12:01 12:02 12:03 12:04 avg latency 0.3 0.5 0.8 1.1 2.6

Slide 14

Slide 14 text

Metrics 12:00 12:01 12:02 12:03 12:04 avg latency 0.3 0.5 0.8 1.1 2.6 server load 0.3 1.0 2.3 3.5 5.2

Slide 15

Slide 15 text

No content

Slide 16

Slide 16 text

Instrument

Slide 17

Slide 17 text

Instrument

Slide 18

Slide 18 text

Instrument

Slide 19

Slide 19 text

Instrument

Slide 20

Slide 20 text

Instrument

Slide 21

Slide 21 text

No content

Slide 22

Slide 22 text

No content

Slide 23

Slide 23 text

Metric Types

Slide 24

Slide 24 text

Metric Types ❖counter

Slide 25

Slide 25 text

Metric Types ❖counter ❖gauge

Slide 26

Slide 26 text

Metric Types ❖counter ❖gauge ❖summary

Slide 27

Slide 27 text

Metric Types ❖counter ❖gauge ❖summary ❖histogram

Slide 28

Slide 28 text

Metric Types ❖counter ❖gauge ❖summary ❖histogram ❖ buckets (1s, 0.5s, 0.25, …)

Slide 29

Slide 29 text

Averages

Slide 30

Slide 30 text

❖ avg(request time) ≠ avg(UX) Averages

Slide 31

Slide 31 text

❖ avg(request time) ≠ avg(UX) ❖ avg({1, 1, 1, 1, 10}) = 2.8 Averages

Slide 32

Slide 32 text

❖ avg(request time) ≠ avg(UX) ❖ avg({1, 1, 1, 1, 10}) = 2.8 Averages

Slide 33

Slide 33 text

❖ avg(request time) ≠ avg(UX) ❖ avg({1, 1, 1, 1, 10}) = 2.8 Averages

Slide 34

Slide 34 text

❖ avg(request time) ≠ avg(UX) ❖ avg({1, 1, 1, 1, 10}) = 2.8 ❖ median({1, 1, 1, 1, 10}) = 1 Averages

Slide 35

Slide 35 text

❖ avg(request time) ≠ avg(UX) ❖ avg({1, 1, 1, 1, 10}) = 2.8 ❖ median({1, 1, 1, 1, 10}) = 1 Averages

Slide 36

Slide 36 text

❖ avg(request time) ≠ avg(UX) ❖ avg({1, 1, 1, 1, 10}) = 2.8 ❖ median({1, 1, 1, 1, 10}) = 1 ❖ median({1, 1, 100_000}) = 1 Averages

Slide 37

Slide 37 text

Percentiles

Slide 38

Slide 38 text

Percentiles nth percentile P of a data set = P ≥ n% of values

Slide 39

Slide 39 text

No content

Slide 40

Slide 40 text

50th percentile = 1 ms

Slide 41

Slide 41 text

50th percentile = 1 ms 50% of requests done by 1 ms

Slide 42

Slide 42 text

Percentiles

Slide 43

Slide 43 text

Percentiles P {1, 1, 100_000} 50th 1

Slide 44

Slide 44 text

Percentiles P {1, 1, 100_000} 50th 1 95th 90_000

Slide 45

Slide 45 text

No content

Slide 46

Slide 46 text

No content

Slide 47

Slide 47 text

No content

Slide 48

Slide 48 text

Naming

Slide 49

Slide 49 text

Naming backend1_app_http_reqs_msgs_post backend1_app_http_reqs_msgs_get …

Slide 50

Slide 50 text

Naming backend1_app_http_reqs_msgs_post backend1_app_http_reqs_msgs_get … app_http_reqs_total

Slide 51

Slide 51 text

Naming backend1_app_http_reqs_msgs_post backend1_app_http_reqs_msgs_get … app_http_reqs_total

Slide 52

Slide 52 text

Naming backend1_app_http_reqs_msgs_post backend1_app_http_reqs_msgs_get … app_http_reqs_total

Slide 53

Slide 53 text

Naming backend1_app_http_reqs_msgs_post backend1_app_http_reqs_msgs_get … app_http_reqs_total{meth="POST", path="/msgs", backend="1"} app_http_reqs_total{meth="GET", path="/msgs", backend="1"} … app_http_reqs_total

Slide 54

Slide 54 text

No content

Slide 55

Slide 55 text

No content

Slide 56

Slide 56 text

1. resolution = scraping interval

Slide 57

Slide 57 text

1. resolution = scraping interval 2. missing scrapes = less resolution

Slide 58

Slide 58 text

Pull: Problems ❖ short lived jobs

Slide 59

Slide 59 text

No content

Slide 60

Slide 60 text

Pull: Problems ❖ short lived jobs ❖ target discovery

Slide 61

Slide 61 text

Configuration scrape_configs: - job_name: 'prometheus' target_groups: - targets: - 'localhost:9090'

Slide 62

Slide 62 text

Configuration scrape_configs: - job_name: 'prometheus' target_groups: - targets: - 'localhost:9090'

Slide 63

Slide 63 text

Configuration scrape_configs: - job_name: 'prometheus' target_groups: - targets: - 'localhost:9090'

Slide 64

Slide 64 text

Configuration scrape_configs: - job_name: 'prometheus' target_groups: - targets: - 'localhost:9090' {instance="localhost:9090",job="prometheus"}

Slide 65

Slide 65 text

No content

Slide 66

Slide 66 text

Pull: Problems ❖ target discovery ❖ short lived jobs ❖ Heroku/NATed systems

Slide 67

Slide 67 text

Pull: Advantages

Slide 68

Slide 68 text

Pull: Advantages ❖ multiple Prometheis easy

Slide 69

Slide 69 text

Pull: Advantages ❖ multiple Prometheis easy ❖ outage detection

Slide 70

Slide 70 text

Pull: Advantages ❖ multiple Prometheis easy ❖ outage detection ❖ predictable, no self-DoS

Slide 71

Slide 71 text

Pull: Advantages ❖ multiple Prometheis easy ❖ outage detection ❖ predictable, no self-DoS ❖ easy to instrument 3rd parties

Slide 72

Slide 72 text

Metrics Format # HELP req_seconds Time spent \ processing a request in seconds. # TYPE req_seconds histogram req_seconds_count 390.0 req_seconds_sum 177.0319407

Slide 73

Slide 73 text

Metrics Format # HELP req_seconds Time spent \ processing a request in seconds. # TYPE req_seconds histogram req_seconds_count 390.0 req_seconds_sum 177.0319407

Slide 74

Slide 74 text

Metrics Format # HELP req_seconds Time spent \ processing a request in seconds. # TYPE req_seconds histogram req_seconds_count 390.0 req_seconds_sum 177.0319407

Slide 75

Slide 75 text

Metrics Format # HELP req_seconds Time spent \ processing a request in seconds. # TYPE req_seconds histogram req_seconds_count 390.0 req_seconds_sum 177.0319407

Slide 76

Slide 76 text

Metrics Format # HELP req_seconds Time spent \ processing a request in seconds. # TYPE req_seconds histogram req_seconds_count 390.0 req_seconds_sum 177.0319407

Slide 77

Slide 77 text

Percentiles req_seconds_bucket{le="0.05"} 0.0 req_seconds_bucket{le="0.25"} 1.0 req_seconds_bucket{le="0.5"} 273.0 req_seconds_bucket{le="0.75"} 369.0 req_seconds_bucket{le="1.0"} 388.0 req_seconds_bucket{le="2.0"} 390.0 req_seconds_bucket{le="+Inf"} 390.0

Slide 78

Slide 78 text

Percentiles req_seconds_bucket{le="0.05"} 0.0 req_seconds_bucket{le="0.25"} 1.0 req_seconds_bucket{le="0.5"} 273.0 req_seconds_bucket{le="0.75"} 369.0 req_seconds_bucket{le="1.0"} 388.0 req_seconds_bucket{le="2.0"} 390.0 req_seconds_bucket{le="+Inf"} 390.0

Slide 79

Slide 79 text

Percentiles req_seconds_bucket{le="0.05"} 0.0 req_seconds_bucket{le="0.25"} 1.0 req_seconds_bucket{le="0.5"} 273.0 req_seconds_bucket{le="0.75"} 369.0 req_seconds_bucket{le="1.0"} 388.0 req_seconds_bucket{le="2.0"} 390.0 req_seconds_bucket{le="+Inf"} 390.0

Slide 80

Slide 80 text

No content

Slide 81

Slide 81 text

Aggregation

Slide 82

Slide 82 text

Aggregation sum( rate( req_seconds_count[1m] ) )

Slide 83

Slide 83 text

Aggregation sum( rate( req_seconds_count[1m] ) )

Slide 84

Slide 84 text

Aggregation sum( rate( req_seconds_count[1m] ) )

Slide 85

Slide 85 text

Aggregation sum( rate( req_seconds_count[1m] ) )

Slide 86

Slide 86 text

Aggregation sum( rate( req_seconds_count{dc="west"}[1m] ) )

Slide 87

Slide 87 text

Aggregation sum( rate( req_seconds_count[1m] ) ) by (dc)

Slide 88

Slide 88 text

Percentiles histogram_quantile( 0.9, rate( req_seconds_bucket[10m] ))

Slide 89

Slide 89 text

Percentiles histogram_quantile( 0.9, rate( req_seconds_bucket[10m] ))

Slide 90

Slide 90 text

Percentiles histogram_quantile( 0.9, rate( req_seconds_bucket[10m] ))

Slide 91

Slide 91 text

Percentiles histogram_quantile( 0.9, rate( req_seconds_bucket[10m] ))

Slide 92

Slide 92 text

Percentiles histogram_quantile( 0.9, rate( req_seconds_bucket[10m] ))

Slide 93

Slide 93 text

No content

Slide 94

Slide 94 text

No content

Slide 95

Slide 95 text

Internal

Slide 96

Slide 96 text

Internal ❖ great for ad-hoc

Slide 97

Slide 97 text

Internal ❖ great for ad-hoc ❖ 1 expr per graph

Slide 98

Slide 98 text

Internal ❖ great for ad-hoc ❖ 1 expr per graph ❖ templating

Slide 99

Slide 99 text

PromDash

Slide 100

Slide 100 text

PromDash ❖ best integration

Slide 101

Slide 101 text

PromDash ❖ best integration ❖ former official

Slide 102

Slide 102 text

PromDash ❖ best integration ❖ former official ❖ now deprecated ❖ don’t bother

Slide 103

Slide 103 text

Grafana

Slide 104

Slide 104 text

Grafana ❖ pretty & powerful

Slide 105

Slide 105 text

Grafana ❖ pretty & powerful ❖ many integrations

Slide 106

Slide 106 text

Grafana ❖ pretty & powerful ❖ many integrations ❖ mix and match!

Slide 107

Slide 107 text

Grafana ❖ pretty & powerful ❖ many integrations ❖ mix and match! ❖ use this!

Slide 108

Slide 108 text

No content

Slide 109

Slide 109 text

Alerts & Scrying

Slide 110

Slide 110 text

Alerts & Scrying ALERT DiskWillFillIn4Hours IF predict_linear( node_filesystem_free[1h], 4*3600) < 0 FOR 5m

Slide 111

Slide 111 text

Alerts & Scrying ALERT DiskWillFillIn4Hours IF predict_linear( node_filesystem_free[1h], 4*3600) < 0 FOR 5m

Slide 112

Slide 112 text

Alerts & Scrying ALERT DiskWillFillIn4Hours IF predict_linear( node_filesystem_free[1h], 4*3600) < 0 FOR 5m

Slide 113

Slide 113 text

Alerts & Scrying ALERT DiskWillFillIn4Hours IF predict_linear( node_filesystem_free[1h], 4*3600) < 0 FOR 5m

Slide 114

Slide 114 text

Alerts & Scrying ALERT DiskWillFillIn4Hours IF predict_linear( node_filesystem_free[1h], 4*3600) < 0 FOR 5m

Slide 115

Slide 115 text

Alerts & Scrying ALERT DiskWillFillIn4Hours IF predict_linear( node_filesystem_free[1h], 4*3600) < 0 FOR 5m

Slide 116

Slide 116 text

No content

Slide 117

Slide 117 text

No content

Slide 118

Slide 118 text

No content

Slide 119

Slide 119 text

Environment

Slide 120

Slide 120 text

No content

Slide 121

Slide 121 text

Apache nginx Django PostgreSQL MySQL MongoDB CouchDB redis Varnish etcd Kubernetes Consul collectd HAProxy statsd graphite InfluxDB SNMP

Slide 122

Slide 122 text

Apache nginx Django PostgreSQL MySQL MongoDB CouchDB redis Varnish etcd Kubernetes Consul collectd HAProxy statsd graphite InfluxDB SNMP

Slide 123

Slide 123 text

node_exporter

Slide 124

Slide 124 text

node_exporter cAdvisor

Slide 125

Slide 125 text

System Insight

Slide 126

Slide 126 text

System Insight ❖ load

Slide 127

Slide 127 text

System Insight ❖ load ❖ procs

Slide 128

Slide 128 text

System Insight ❖ load ❖ procs ❖ memory

Slide 129

Slide 129 text

System Insight ❖ load ❖ procs ❖ memory ❖ network

Slide 130

Slide 130 text

System Insight ❖ load ❖ procs ❖ memory ❖ network ❖ disk

Slide 131

Slide 131 text

System Insight ❖ load ❖ procs ❖ memory ❖ network ❖ disk ❖ I/O

Slide 132

Slide 132 text

mtail

Slide 133

Slide 133 text

mtail ❖ follow (log) files

Slide 134

Slide 134 text

mtail ❖ follow (log) files ❖ extract metrics using regex

Slide 135

Slide 135 text

mtail ❖ follow (log) files ❖ extract metrics using regex ❖ can be better than direct

Slide 136

Slide 136 text

Moar

Slide 137

Slide 137 text

Moar ❖ Edges: web servers/HAProxy

Slide 138

Slide 138 text

Moar ❖ Edges: web servers/HAProxy ❖ black box

Slide 139

Slide 139 text

Moar ❖ Edges: web servers/HAProxy ❖ black box ❖ databases

Slide 140

Slide 140 text

Moar ❖ Edges: web servers/HAProxy ❖ black box ❖ databases ❖ network

Slide 141

Slide 141 text

So Far

Slide 142

Slide 142 text

So Far ❖ system stats

Slide 143

Slide 143 text

So Far ❖ system stats ❖ outside look

Slide 144

Slide 144 text

So Far ❖ system stats ❖ outside look ❖ 3rd party components

Slide 145

Slide 145 text

Code

Slide 146

Slide 146 text

cat-or.not

Slide 147

Slide 147 text

cat-or.not ❖ HTTP service

Slide 148

Slide 148 text

cat-or.not ❖ HTTP service ❖ upload picture

Slide 149

Slide 149 text

cat-or.not ❖ HTTP service ❖ upload picture ❖ meow!/nope meow!

Slide 150

Slide 150 text

from flask import Flask, g, request from cat_or_not import is_cat app = Flask(__name__) @app.route("/analyze", methods=["POST"]) def analyze(): g.auth.check(request) return ("meow!" if is_cat(request.files["pic"]) else "nope!") if __name__ == "__main__": app.run()

Slide 151

Slide 151 text

from flask import Flask, g, request from cat_or_not import is_cat app = Flask(__name__) @app.route("/analyze", methods=["POST"]) def analyze(): g.auth.check(request) return ("meow!" if is_cat(request.files["pic"]) else "nope!") if __name__ == "__main__": app.run()

Slide 152

Slide 152 text

from flask import Flask, g, request from cat_or_not import is_cat app = Flask(__name__) @app.route("/analyze", methods=["POST"]) def analyze(): g.auth.check(request) return ("meow!" if is_cat(request.files["pic"]) else "nope!") if __name__ == "__main__": app.run()

Slide 153

Slide 153 text

pip install prometheus_client

Slide 154

Slide 154 text

from prometheus_client import \ start_http_server # … if __name__ == "__main__": start_http_server(8000) app.run()

Slide 155

Slide 155 text

process_virtual_memory_bytes 156393472.0 process_resident_memory_bytes 20480000.0 process_start_time_seconds 1460214325.21 process_cpu_seconds_total 0.169999999998 process_open_fds 8.0 process_max_fds 1024.0

Slide 156

Slide 156 text

process_virtual_memory_bytes 156393472.0 process_resident_memory_bytes 20480000.0 process_start_time_seconds 1460214325.21 process_cpu_seconds_total 0.169999999998 process_open_fds 8.0 process_max_fds 1024.0

Slide 157

Slide 157 text

process_virtual_memory_bytes 156393472.0 process_resident_memory_bytes 20480000.0 process_start_time_seconds 1460214325.21 process_cpu_seconds_total 0.169999999998 process_open_fds 8.0 process_max_fds 1024.0

Slide 158

Slide 158 text

process_virtual_memory_bytes 156393472.0 process_resident_memory_bytes 20480000.0 process_start_time_seconds 1460214325.21 process_cpu_seconds_total 0.169999999998 process_open_fds 8.0 process_max_fds 1024.0

Slide 159

Slide 159 text

process_virtual_memory_bytes 156393472.0 process_resident_memory_bytes 20480000.0 process_start_time_seconds 1460214325.21 process_cpu_seconds_total 0.169999999998 process_open_fds 8.0 process_max_fds 1024.0

Slide 160

Slide 160 text

process_virtual_memory_bytes 156393472.0 process_resident_memory_bytes 20480000.0 process_start_time_seconds 1460214325.21 process_cpu_seconds_total 0.169999999998 process_open_fds 8.0 process_max_fds 1024.0

Slide 161

Slide 161 text

No content

Slide 162

Slide 162 text

from prometheus_client import \ Histogram, Gauge REQUEST_TIME = Histogram( "cat_or_not_request_seconds", "Time spent in HTTP requests.")

Slide 163

Slide 163 text

from prometheus_client import \ Histogram, Gauge REQUEST_TIME = Histogram( "cat_or_not_request_seconds", "Time spent in HTTP requests.") ANALYZE_TIME = Histogram( "cat_or_not_analyze_seconds", "Time spent analyzing pictures.")

Slide 164

Slide 164 text

from prometheus_client import \ Histogram, Gauge REQUEST_TIME = Histogram( "cat_or_not_request_seconds", "Time spent in HTTP requests.") ANALYZE_TIME = Histogram( "cat_or_not_analyze_seconds", "Time spent analyzing pictures.") IN_PROGRESS = Gauge( "cat_or_not_in_progress_requests", "Number of requests in progress.")

Slide 165

Slide 165 text

@IN_PROGRESS.track_inprogress() @REQUEST_TIME.time() @app.route("/analyze", methods=["POST"]) def analyze(): g.auth.check(request) with ANALYZE_TIME.time(): result = is_cat( request.files["pic"].stream) return "meow!" if result else "nope!"

Slide 166

Slide 166 text

@IN_PROGRESS.track_inprogress() @REQUEST_TIME.time() @app.route("/analyze", methods=["POST"]) def analyze(): g.auth.check(request) with ANALYZE_TIME.time(): result = is_cat( request.files["pic"].stream) return "meow!" if result else "nope!"

Slide 167

Slide 167 text

AUTH_TIME = Histogram("auth_seconds", "Time spent authenticating.") AUTH_ERRS = Counter("auth_errors_total", "Errors while authing.") AUTH_WRONG_CREDS = Counter("auth_wrong_creds_total", "Wrong credentials.") class Auth: # ... @AUTH_TIME.time() def auth(self, request): while True: try: return self._auth(request) except WrongCredsError: AUTH_WRONG_CREDS.inc() raise except Exception: AUTH_ERRS.inc()

Slide 168

Slide 168 text

AUTH_TIME = Histogram("auth_seconds", "Time spent authenticating.") AUTH_ERRS = Counter("auth_errors_total", "Errors while authing.") AUTH_WRONG_CREDS = Counter("auth_wrong_creds_total", "Wrong credentials.") class Auth: # ... @AUTH_TIME.time() def auth(self, request): while True: try: return self._auth(request) except WrongCredsError: AUTH_WRONG_CREDS.inc() raise except Exception: AUTH_ERRS.inc()

Slide 169

Slide 169 text

AUTH_TIME = Histogram("auth_seconds", "Time spent authenticating.") AUTH_ERRS = Counter("auth_errors_total", "Errors while authing.") AUTH_WRONG_CREDS = Counter("auth_wrong_creds_total", "Wrong credentials.") class Auth: # ... @AUTH_TIME.time() def auth(self, request): while True: try: return self._auth(request) except WrongCredsError: AUTH_WRONG_CREDS.inc() raise except Exception: AUTH_ERRS.inc()

Slide 170

Slide 170 text

AUTH_TIME = Histogram("auth_seconds", "Time spent authenticating.") AUTH_ERRS = Counter("auth_errors_total", "Errors while authing.") AUTH_WRONG_CREDS = Counter("auth_wrong_creds_total", "Wrong credentials.") class Auth: # ... @AUTH_TIME.time() def auth(self, request): while True: try: return self._auth(request) except WrongCredsError: AUTH_WRONG_CREDS.inc() raise except Exception: AUTH_ERRS.inc()

Slide 171

Slide 171 text

@app.route("/analyze", methods=["POST"]) def analyze(): g.auth.check(request) with ANALYZE_TIME.time(): result = is_cat( request.files["pic"].stream) return "meow!" if result else "nope!"

Slide 172

Slide 172 text

pip install prometheus_async

Slide 173

Slide 173 text

Wrapper from prometheus_async.aio import time @time(REQUEST_TIME) async def view(request): # ...

Slide 174

Slide 174 text

Goodies

Slide 175

Slide 175 text

Goodies ❖ aiohttp-based metrics export

Slide 176

Slide 176 text

Goodies ❖ aiohttp-based metrics export ❖ also in thread!

Slide 177

Slide 177 text

Goodies ❖ aiohttp-based metrics export ❖ also in thread! ❖ Consul Agent integration

Slide 178

Slide 178 text

Wrap Up

Slide 179

Slide 179 text

Wrap Up

Slide 180

Slide 180 text

Wrap Up ✓

Slide 181

Slide 181 text

Wrap Up ✓ ✓

Slide 182

Slide 182 text

Wrap Up ✓ ✓ ✓

Slide 183

Slide 183 text

ox.cx/p @hynek vrmd.de