1 of 44

After six years, is GitOps boring?

Alexis Richardson, CEO Weaveworks

GitOpsConEU Keynote, 5 December 2023

2 of 44

2

3 of 44

3

2023

4 of 44

4

Automation =

GitOps +

“GitOps is now boring”

5 of 44

100% adoption is BORING

5

Read our summary of the new CNCF Microsurvey on GitOps adoption!

https://www.cncf.io/blog/2023/12/01/gitops-goes-mainstream-flux-cd-boasts-largest-ecosystem/

6 of 44

6

Automation =

GitOps +

so what is exciting?

7 of 44

7

AI is exciting!

8 of 44

8

The Developer Experience Gap

Automation is the answer

But how do we scale GitOps to deliver Automation?

Platforms!

Can AI make better Platforms?

9 of 44

2017

GitOps

what will 100M developers do?

what about 1B?

what tools will they use?

→ “the DX gap”

Please credit weaveworks if you share

9

10 of 44

| 10

Tech

Overload

Change Acceleration

Management Complexity

Increasing Demand

Tighter

Budgets

Low tech productivity → High CIO pain

11 of 44

11

Automation

12 of 44

gitops automation means that

Developers can safely and easily deploy and operate more apps

12

13 of 44

scaling to many teams..

  • most tech people understand apps, not kubernetes, terraform, yaml
  • devs want frictionless delivery, with security and compliance
  • ops want fleets and SLOs but without adding config or other burdens

See also GitOps at Scale presentation @ March 23

13

14 of 44

CIOs want:

Best practices for teams of developers at scale

15 of 44

15

Platforms

16 of 44

16

This is a Boring Platform

17 of 44

GOALS

10x developer productivity

90% reduction in operations costs

achieving platform scale is not boring!

17

18 of 44

18

platform

ops

dev

“Platform” is an automation contract

simplify applications for all developer teams by creating instant environments for them to run their applications safely and without ops pain

19 of 44

19

platform

ops

dev

“Platform” is an automation contract

Enforcing

  • Simple Integration
  • Fast Configuration
  • Trust, Security and Support
  • Scale Compliance for all Teams

20 of 44

Understand, Deploy and Operate Cloud Native Applications

TEAM

Onboard app and platform teams quickly and safely

CATALOG

Document and share templates, dependencies, policies, addons

KUBERNETES

Simplify Kubernetes on any cluster lifecycle manager

PIPELINES

Deploy, stage, promote and verify rollouts of apps and infra

SECURITY

Trusted application delivery with signed and verified assets

TERRAFORM+

Coordinate rollouts with TF other infra-as-code tools

FLEETS

Scale up operations for applications and infrastructure

COMPLIANCE

Automated safety alerts, validation, policy management

DATA and AI

Integrate external data and AI workflows into the platform

ANALYTICS

Test, report and act on KPIs and SLOs for devops and finops

OBSERVABILITY

Add standard dashboards and tools for alerting, logging etc

CONNECT

Configure and manage API gateways, networks, routing

weaveworks

21 of 44

21

Examples

22 of 44

22

23 of 44

23

24 of 44

24

Harnessing Argo & Flux: The Quest to Scale Add-Ons Beyond 10,000 Clusters- Joaquin Rodriguez, Microsoft & Priyanka Ravi, Weaveworks

https://www.youtube.com/watch?v=IPySUmNKbcA

25 of 44

25

Weave AI Controllers: GitOps Automation for Large Language Models (LLMs)

https://www.weave.works/blog/weave-ai-controllers

26 of 44

26

27 of 44

27

BORING

28 of 44

28

AI

29 of 44

29

Get help with writing the YAMLs

It looks like you’re using Kubernetes .

Would you like help?

Just run the damn thing for me now

Don’t show me this tip again

GitOps + AI?

30 of 44

30

GitOps + AI = Automation Platform

31 of 44

31

32 of 44

32

33 of 44

33

34 of 44

34

Is English the best programming language?

35 of 44

35

GENERATIVE GitOps

36 of 44

36

platform

ops

dev

INTENT

make code and yaml for me

37 of 44

37

platform

ops

dev

EXPLAIN

ANOMALY!

38 of 44

Jobs

Workflows

Imperative

Declarative

Continuous Integration tools run arbitrary jobs.

Abstract deployment tasks to create stand-alone workflow building blocks.

Platform supports actions and deployment tasks across the entire SDLC process, from build to multiple deploy targets.

Platform defines end state and compliance with best practices, and automated GitOps agents enforce correct deployment and drift management.

Towards a Generative Automation Platform

Generative

Users ask questions like “how can I improve throughput?” and the Platform AI reconfigures the system to achieve this

Legacy

Modern

Future

credit adam frank, armory

39 of 44

39

platform

ops

dev

DEVELOPER INTENT

platform

ops

dev

AUTOMATE OPERATIONS

PLATFORM

EXPRESSES

& MANAGES DESIRED STATE

40 of 44

40

Can feedback improve performance and remove LLM hallucinations?

41 of 44

41

42 of 44

42

2024?

43 of 44

2024 will be boring

44 of 44

Thank You!

Alexis Richardson

@monadic still on twitter for now