KubeCon Europe 2018: Switching Horses Midstream: The Challenges of Migrating 150+ Microservices to Kubernetes

Slide 1

Slide 1 text

Switching horses midstream: the challenge of migrating 150+ services to kubernetes Sarah Wells Technical Director for Operations and Reliability, Financial Times @sarahjwells

Slide 2

Slide 2 text

No content

Slide 3

Slide 3 text

No content

Slide 4

Slide 4 text

No content

Slide 5

Slide 5 text

No content

Slide 6

Slide 6 text

The FT’s Content platform

Slide 7

Slide 7 text

This is what it really looks like…

Slide 8

Slide 8 text

@sarahjwells Why *did* we migrate to k8s?

Slide 9

Slide 9 text

@sarahjwells Mid 2015: a hand-rolled container stack

Slide 10

Slide 10 text

@sarahjwells https://medium.com/wardleymaps

Slide 11

Slide 11 text

@sarahjwells Spend your innovation tokens wisely

Slide 12

Slide 12 text

@sarahjwells ~80% reduction in EC2 costs

Slide 13

Slide 13 text

@sarahjwells Many fewer steps to start running a new service in production

Slide 14

Slide 14 text

@sarahjwells But: supportability of an in-house platform is a challenge

Slide 15

Slide 15 text

@sarahjwells http://mcfunley.com/choose-boring-technology Choose boring technology

Slide 16

Slide 16 text

@sarahjwells By late 2016, tools were maturing

Slide 17

Slide 17 text

@sarahjwells https://medium.com/wardleymaps

Slide 18

Slide 18 text

@sarahjwells The FT is not a cluster orchestration company

Slide 19

Slide 19 text

@sarahjwells Late 2016: Consider the alternatives

Slide 20

Slide 20 text

@sarahjwells Metrics for success: - amount of time spent keeping cluster healthy - number of sarcastic comments on slack

Slide 21

Slide 21 text

@sarahjwells Opted for kubernetes

Slide 22

Slide 22 text

No content

Slide 23

Slide 23 text

@sarahjwells Using leading edge technologies requires you to be comfortable with change

Slide 24

Slide 24 text

@sarahjwells Shouldn’t be (too) scared about making the wrong decision http://uk.businessinsider.com/jeff-bezos-on-type-1-and-type-2- decisions-2016-4

Slide 25

Slide 25 text

@sarahjwells Switching horses midstream

Slide 26

Slide 26 text

@sarahjwells At the start of this migration we had 150 services

Slide 27

Slide 27 text

@sarahjwells Lots of other work going on at the same time

Slide 28

Slide 28 text

@sarahjwells Complications of running in parallel

Slide 29

Slide 29 text

@sarahjwells We had well over 2000 code releases while running at least part of the stack in parallel

Slide 30

Slide 30 text

@sarahjwells Decisions, decisions, decisions…

Slide 31

Slide 31 text

@sarahjwells Separate branches vs if/else in code

Slide 32

Slide 32 text

@sarahjwells Separate deployment mechanisms vs a single deployment mechanism

Slide 33

Slide 33 text

@sarahjwells Risk-based approach to testing

Slide 34

Slide 34 text

@sarahjwells Doing anything 150 times takes time

Slide 35

Slide 35 text

@sarahjwells Changes per service weren’t *that* big

Slide 36

Slide 36 text

@sarahjwells Migrating from systemd service ﬁles to helm charts

Slide 37

Slide 37 text

@sarahjwells Integrating the service into a templated jenkins pipeline

Slide 38

Slide 38 text

@sarahjwells Good to get everyone involved - “Helm days”

Slide 39

Slide 39 text

@sarahjwells Discovered a lot of ‘broken’ things

Slide 40

Slide 40 text

@sarahjwells Services that hadn’t been built for a long time

Slide 41

Slide 41 text

@sarahjwells A standard that isn’t enforced may will not be complied with: - healthcheck timeouts

Slide 42

Slide 42 text

@sarahjwells - /__gtg endpoints

Slide 43

Slide 43 text

@sarahjwells Making sure a service will recover if k8s moves it elsewhere

Slide 44

Slide 44 text

@sarahjwells Easy to get sucked into making things better

Slide 45

Slide 45 text

@sarahjwells Would have been better if…

Slide 46

Slide 46 text

@sarahjwells We’d swarmed on the work

Slide 47

Slide 47 text

@sarahjwells The longer you run in parallel, the more overhead for releasing code changes