SRE Made Easy

Have you ever had to set up telemetry for a bunch of microservices and found your self surrounded by metrics with no observability in sight? The GCP Service Mesh promises a “batteries included” experience for observability and security. This talk aims to challenge that – we will look at how easy it is to set up, what you get out of the box and how quickly you reach the limits. We will also compare the tool with its open-source counterpart, Istio.

Transcript

2. Do it RIGHT. What is SRE
3. Do it RIGHT. Site Reliability Engineering 3 Culture • Customer focus, measure the impact on the user instead of system metrics • Empower teams to call for a code freeze and focus on reliability when needed • Accept errors as a fact of life, find the appropriate level of reliability and practice recovery
4. Do it RIGHT. Site Reliability Engineering 4 Technology • Automate things • Follow the reliability and scalability architectural patterns • Make it easy to do the right thing and difficult to do the wrong thing
5. Do it RIGHT. Site Reliability Engineering 5 Observability • When you have 3 servers you can check up on them from time to time and be the human monitor • When you have 3 000 servers you need a machine monitor • If you don’t have observability, it just means your users are the observability stack
6. Do it RIGHT. Site Reliability Engineering 6
7. Do it RIGHT. What is a Service Mesh
8. Do it RIGHT. Service Mesh 8 Routing • Service discovery and load balancing • Layer 7 protocol-aware routing • Traffic splitting, A/B testing • External services with retries, timeouts, fault injection
9. Do it RIGHT. Service Mesh 9 Security • Authentication and Authorization setup • Certificates & mTLS • Rate limiting, Circuit breakers
10. Do it RIGHT. Service Mesh 10 Observability • Logging • Monitoring • Distributed tracing • Alerting
11. Do it RIGHT. How it all comes together
12. Do it RIGHT. Time for a demo • Set up a demo microservices project on GKE • Use the GCP Service Mesh • Use the GCP Observability stack for SLOs • Try a load test
13. Do it RIGHT. Where it falls short
14. Do it RIGHT. Standard metrics • Golden signals are sometimes not enough • Remember the customer focus – the user defines reliability • Some instrumentation is still needed
15. Do it RIGHT. THANK YOU! CONTACT US: Weyringergasse 1-3/DG 1040 Wien www.posedio.com office@posedio.com 15