Tracing the Black Box of Serverless

A presentation at Abstractions II in August 2019 in Pittsburgh, PA, USA by Matt Williams

Slide 1

Tracing the Black Box of Serverless matt williams – evangelist @ datadog

Slide 2

5 minutes after we first met

Slide 3

Slide 4

Slide 5

Slide 6

Slide 7

Slide 8

Slide 9

Slide 10

Slide 11

Slide 12

Slide 13

How I use this

Slide 14

Slide 15

Slide 16

It works great

Slide 17

No errors… Data shows up where its supposed to

Slide 18

But it feels slow

Slide 19

How do I find out where the problem is…

Slide 20

…when there is no problem

Slide 21

Three pillars of observability • Metrics • Spotting trends/patterns • Send alerts • Catch known unknowns • Logs • Traces

Slide 22

Slide 23

Slide 24

Slide 25

Relevant metrics to help me • Total ounces eaten • Performance timings for each function

Slide 26

Three pillars of observability • Metrics • Logs • Structured & easy to grep • Find the details of an event • Catch the unknown unknowns • Traces

Slide 27

Slide 28

Slide 29

Slide 30

Relevant log lines for me • Function entry and exit timestamps, total time • What is being collected by each step

Slide 31

Three pillars of observability • Metrics • Logs • Traces • Request-based • Follow activity across function and service • Answers “Where?” and “How long?”

Slide 32

Slide 33

Slide 34

Slide 35

Slide 36

Slide 37

Slide 38

But I want more visibility into what is going on

Slide 39

Slide 40

Slide 41

Slide 42

Time wasted getting static values

Slide 43

Slide 44

Slide 45

But I need more data

Slide 46

Slide 47

Slide 48

Slide 49

Slide 50

Matt Williams Evangelist at Datadog @technovangelist mattw@datadoghq.com youtube.com/technovangelist