Slide 1

Slide 1 text

5 Years of Metrics & Monitoring Lindsay Holmwood @auxesis

Slide 2

Slide 2 text

Cultural & Technical

Slide 3

Slide 3 text

• Key retrospective questions • What did we do well? • What did we learn? • What should we do differently next time? • What still puzzles us?

Slide 4

Slide 4 text

What got us here won’t get us there

Slide 5

Slide 5 text

What did we do well? (that if we don’t talk about, we might forget)

Slide 6

Slide 6 text

The Pipeline

Slide 7

Slide 7 text

No content

Slide 8

Slide 8 text

collection

Slide 9

Slide 9 text

storage collection

Slide 10

Slide 10 text

storage checking collection

Slide 11

Slide 11 text

storage checking alerting collection

Slide 12

Slide 12 text

storage checking alerting collection graphing

Slide 13

Slide 13 text

storage checking alerting collection graphing aggregation

Slide 14

Slide 14 text

No content

Slide 15

Slide 15 text

collection storage checking alerting graphing aggregation

Slide 16

Slide 16 text

collection storage checking alerting graphing aggregation collectd & statsd

Slide 17

Slide 17 text

collection storage checking alerting graphing aggregation Graphite & OpenTSDB & InfluxDB

Slide 18

Slide 18 text

collection storage checking alerting graphing aggregation Riemann

Slide 19

Slide 19 text

Alert fatigue has become a recognised problem

Slide 20

Slide 20 text

Cottage industry

Slide 21

Slide 21 text

PagerDuty & VictorOps & OpsGenie

Slide 22

Slide 22 text

#monitoringsucks

Slide 23

Slide 23 text

#monitoringlove

Slide 24

Slide 24 text

No content

Slide 25

Slide 25 text

What would we do differently next time?

Slide 26

Slide 26 text

Graphs & Dashboards

Slide 27

Slide 27 text

Apparently the hardest problem in monitoring is graphing and dashboarding.

Slide 28

Slide 28 text

What we’re doing wrong

Slide 29

Slide 29 text

Strip charts

Slide 30

Slide 30 text

No content

Slide 31

Slide 31 text

No content

Slide 32

Slide 32 text

No content

Slide 33

Slide 33 text

We have a problem

Slide 34

Slide 34 text

Strip charts: the PHP hammer of graphing

Slide 35

Slide 35 text

What can the data tell us?

Slide 36

Slide 36 text

What is the distribution?

Slide 37

Slide 37 text

It’s not a problem with the tools

Slide 38

Slide 38 text

Our approach is tainted

Slide 39

Slide 39 text

graphing problems we have graphing problems serviced by strip charts

Slide 40

Slide 40 text

No content

Slide 41

Slide 41 text

Basic graph layout

Slide 42

Slide 42 text

Black on white

Slide 43

Slide 43 text

bounding box with x + y axes labels 1 2 3 4 5 5 3 1 5 3 1 1 2 3 4 5

Slide 44

Slide 44 text

Colour

Slide 45

Slide 45 text

Differential colour engine

Slide 46

Slide 46 text

No content

Slide 47

Slide 47 text

Maximum of 15 colours on-screen

Slide 48

Slide 48 text

8%

Slide 49

Slide 49 text

Adjust saturation, not hue

Slide 50

Slide 50 text

Use minimal hue to call out data

Slide 51

Slide 51 text

No content

Slide 52

Slide 52 text

Fucking Pie Charts

Slide 53

Slide 53 text

No content

Slide 54

Slide 54 text

Experiment: Compare segment sizes

Slide 55

Slide 55 text

No content

Slide 56

Slide 56 text

No content

Slide 57

Slide 57 text

No content

Slide 58

Slide 58 text

No content

Slide 59

Slide 59 text

– William S. Cleveland, p.86 Principles of Graphing Data This allows us to see very clearly that the pie chart judgements are less accurate than the bar chart judgements.

Slide 60

Slide 60 text

Pie chart comparisons are more error prone

Slide 61

Slide 61 text

Pie not eaten Pie eaten The only time you should use a pie chart

Slide 62

Slide 62 text

No content

Slide 63

Slide 63 text

What did we learn?

Slide 64

Slide 64 text

Democratisation of graphing tool development

Slide 65

Slide 65 text

Scratch our itches

Slide 66

Slide 66 text

Same poor UX, better paint job

Slide 67

Slide 67 text

No content

Slide 68

Slide 68 text

No content

Slide 69

Slide 69 text

We get the graphing tools we deserve

Slide 70

Slide 70 text

Nagios is here to stay (at least for ops)

Slide 71

Slide 71 text

Inertia

Slide 72

Slide 72 text

No strong, compelling alternative

Slide 73

Slide 73 text

Sensu

Slide 74

Slide 74 text

When I hear people say “I'm not using Sensu because it's too complex” I think “and Nagios isn't hiding the same complexity from you?”

Slide 75

Slide 75 text

This is a problem

Slide 76

Slide 76 text

No content

Slide 77

Slide 77 text

We don’t know stats

Slide 78

Slide 78 text

storage checking alerting collection graphing aggregation

Slide 79

Slide 79 text

storage checking alerting collection graphing aggregation checks

Slide 80

Slide 80 text

Numbers & Strings & Behaviour

Slide 81

Slide 81 text

Numbers

Slide 82

Slide 82 text

Fault detection (thresholding)

Slide 83

Slide 83 text

Anomaly detection (trend analysis)

Slide 84

Slide 84 text

Monitoring is CI for Production

Slide 85

Slide 85 text

Continuous Integration

Slide 86

Slide 86 text

1. checkout Continuous Integration

Slide 87

Slide 87 text

1. checkout 2. build Continuous Integration

Slide 88

Slide 88 text

1. checkout 2. build 3. test Continuous Integration

Slide 89

Slide 89 text

1. checkout 2. build 3. test 4. notify Continuous Integration

Slide 90

Slide 90 text

1. checkout 2. build 3. test 4. notify Continuous Integration Monitoring

Slide 91

Slide 91 text

1. checkout 2. build 3. test 4. notify can I see my app? Continuous Integration Monitoring

Slide 92

Slide 92 text

• serverspec & • sensu

Slide 93

Slide 93 text

No content

Slide 94

Slide 94 text

What still puzzles us? (or, what might the future look like?)

Slide 95

Slide 95 text

The future is analysing & acting on our alert data

Slide 96

Slide 96 text

• Last 5 years • Building new tools • Formalising relationships • Search for parallels in other industries • Measuring the human impact

Slide 97

Slide 97 text

• Next • Stabilisation of tools • Emerging standards • Exploiting parallels • Mitigating the human impact

Slide 98

Slide 98 text

Analysis: Ops Weekly

Slide 99

Slide 99 text

No content

Slide 100

Slide 100 text

No content

Slide 101

Slide 101 text

Context: Nagios Herald

Slide 102

Slide 102 text

No content

Slide 103

Slide 103 text

No content

Slide 104

Slide 104 text

The future is richer metadata about our metrics

Slide 105

Slide 105 text

Metrics 2.0

Slide 106

Slide 106 text

{ server: dfs1 what: diskspace mountpoint: srv/node/dfs10 unit: B type: used metric_type: gauge } meta: { agent: diamond, processed_by: statsd2 }

Slide 107

Slide 107 text

Self-describing

Slide 108

Slide 108 text

The future is richer metadata about our metrics

Slide 109

Slide 109 text

The future is richer metadata about our metrics to automatically build appropriate visualisations

Slide 110

Slide 110 text

• Aggregation & • Grouping & • Unit conversions & • Scaling & • Axes labelling & • …

Slide 111

Slide 111 text

Death to strip charts

Slide 112

Slide 112 text

The future is monitoring tools for devs

Slide 113

Slide 113 text

Ops must be enablers, not gatekeepers

Slide 114

Slide 114 text

What has made sense about ops being gatekeepers?

Slide 115

Slide 115 text

Monitoring is treated as an operational responsibility

Slide 116

Slide 116 text

Ops team own ops

Slide 117

Slide 117 text

We’ve won the battles

Slide 118

Slide 118 text

Ops team own ops

Slide 119

Slide 119 text

This is no longer the world we live in

Slide 120

Slide 120 text

How do we become enablers?

Slide 121

Slide 121 text

Technical & Cultural

Slide 122

Slide 122 text

No content

Slide 123

Slide 123 text

• Technical

Slide 124

Slide 124 text

• Technical • Ops provide the platform

Slide 125

Slide 125 text

• Technical • Ops provide the platform • Maintain, monitor, and scale the platform

Slide 126

Slide 126 text

— Adrian Cockcroft

Slide 127

Slide 127 text

No content

Slide 128

Slide 128 text

• Cultural

Slide 129

Slide 129 text

• Cultural • Coach on what makes a good check • Coach on what is good alert design • Listen to the needs of the end-user

Slide 130

Slide 130 text

Provide monitoring as a service

Slide 131

Slide 131 text

Monitoring is a core deliverable on every service

Slide 132

Slide 132 text

Ship checks & config with your applications

Slide 133

Slide 133 text

Example: Yelp

Slide 134

Slide 134 text

No content

Slide 135

Slide 135 text

What’s the barrier to entry?

Slide 136

Slide 136 text

Does the idea just not have traction?

Slide 137

Slide 137 text

Are the tools not up to scratch?

Slide 138

Slide 138 text

Does monitoring need to be SaaS (or SaaS-like) to make this achievable at scale?

Slide 139

Slide 139 text

– William Gibson The future is here – it’s just not very evenly distributed

Slide 140

Slide 140 text

Monitoring is still insular

Slide 141

Slide 141 text

We’re building tools for operations teams

Slide 142

Slide 142 text

Not the developers who need them most

Slide 143

Slide 143 text

No content

Slide 144

Slide 144 text

Monitoring is like a joke.

Slide 145

Slide 145 text

Monitoring is like a joke. If you have to explain it, it’s not that good.

Slide 146

Slide 146 text

storage checking alerting collection graphing aggregation

Slide 147

Slide 147 text

What can we do better?

Slide 148

Slide 148 text

I’m Lindsay @auxesis

Slide 149

Slide 149 text

Dank je wel!

Slide 150

Slide 150 text

Dank je wel! Liked the talk? Let @auxesis know.