monitoring kubernetes vs. serverless · monitoring kubernetes vs. serverless erez berkner...

Post on 17-Aug-2020

19 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Monitoring Kubernetes vs. Serverless

Erez Berkner

Co-Founder & CEO, Lumigo

erez@lumigo.io

@erezberkner

self.getInfo( )

2

Erez Berkner Co-Founder & CEO, Lumigo

Micro-services monitoring platform

Developer

Cloud R&D Manager

Micro-services focus

3

Agenda

• Why Monitor?

• Evolution of Cloud Infrastructure

• Monitoring K8s

• Serverless Changes Everything

• Monitoring Serverless

Why Monitor?

4

Know when something is wrong

Understand business impacts

Find the root cause. Fast.

The Evolution of Cloud Infrastructure

5

The Evolution of Transportation

6

Ownership

Fuel

Navigate

To get there! Focus on

getting there!

The Evolution of Cloud Infrastructure

7

Focus on

business logic!

Physical

Servers

Virtual

Machines Containers Serverless

HW Ownership

OS

Runtime & Scale

Code

Monitoring Kubernetes Best Practices

8

The 5 Layers Of K8s

9

K8s & Prometheus – Best Friends

10

Monitoring K8s – Best Practices [Metrics]

11

Monitoring K8s – Best Practices [Metrics]

12

Monitoring K8s – Best Practices [Logs]

13

Monitoring K8s – Best Practices [Traffic]

14

Monitoring K8s – Service Mesh

15

The new kid …

16

Serverless Impact

17

Distributed Nano Services

Self Managed

Cost Paradigm Change

Async

Dynamic agile env

Serverless Challenges

18

Visibility Understanding cost

Identify & Resolve Issues In a

Distributed Architecture

19

Monitoring in the dark…

20

How do we Monitor Serverless?

Option 1

Cloudwatch

& Friends

Option 2

Homebrewed

Solutions

Option 3

Serverless

Monitoring

Solutions

Design

Option 1

Cloudwatch & Friends

21

Option 1 Cloudwatch & Friends

22

Cons:

• Complicated and time consuming

• No event correlation / tracing

• Hard to understand business

impact

Pros:

• Out of the box

• AWS support

Design

Option 2

Homebrewed Solutions

23

Option 2 Homebrewed Solutions

24

• Generate your own correlation ID

• Pass the correlation ID for every service

• Add the correlation ID for every log

Option 2 Homebrewed Solutions

25

• Use Distributed Tracing frameworks:

o Zipkin (+Jaeger)

o OpenTracing

o OpenCensus

Pros:

• Tailor fit

• Supported by many vendors

• Not cloud specific

Cons:

• Very high-touch

• Managed services (like S3)

• Some components aren’t included (e.g. API GW)

Design

Option 3

Serverless

Monitoring

Solutions

26

27

• SaaS platforms

• Solving variety of challanges

• Common implementation:

• Library

• IAM role

Option 3 Serverless Monitoring Solutions

Pros:

• Serverless focused

• More than just tracing

• Very low touch

Cons:

• Yet another 3rd platform

• More than just tracing

28

DEMO TIME !

29

Takeaways

• 5 layers of K8s – monitor all of them

• Micro services requires Distributed Tracing

• Emerging monitoring challenge around tracing distributed managed services

• Use existing frameworks / tools when available

• Serverless leads to Managed Distributed Tracing

Questions?

30

erez@lumigo.io

Erez Berkner

Co-Founder & CEO, Lumigo

@erezberkner

THANK YOU !

top related