DevOps & Cloud
SRE and CloudOps on AWS
Define reliability, measure it, operate it, break it on purpose, and improve - SLOs and error budgets, instrumentation, alerting, autoscaling, chaos engineering, incident command, multi-AZ and multi-region, postmortems.
SRE and CloudOps on AWS
Reliability is not a feeling. It is a number you agreed to, a budget you spend, and a set of practices for when you have spent it.
The course is a loop rather than a line - define reliability, measure it, operate it, break it, improve - and module 5 feeds straight back into module 1, which is the actual shape of the discipline:
- Reliability foundations - what SRE means on AWS specifically, then SLIs, SLOs and error budgets built on CloudWatch rather than on a whiteboard
- Instrumentation - what to keep when you instrument a distributed system, and alerting that does not destroy your on-call rotation
- Operating at scale - runbooks and Systems Manager Automation, capacity planning and autoscaling that actually works, and load testing that does not lie to you
- Failure - failure injection with AWS FIS, incident command in the first fifteen minutes, what a multi-AZ failure really looks like, and the honest multi-region decision
- Improvement - postmortems against an AWS-specific failure taxonomy, and the automation ladder for reducing toil
Two ideas keep returning. Most reliability work is about deciding what you will not do - an error budget is permission to stop. And most AWS outages are not the failure you rehearsed: the AZ does not go down, it goes slow.
Module 1 is free to preview. The full course is a one-time purchase with lifetime access.
What you will learn
- Write SLIs and SLOs that mean something, and run an error budget you actually spend
- Instrument a distributed system without paying for telemetry nobody reads
- Build alerting that wakes someone only when a human decision is needed
- Automate operations with runbooks and Systems Manager instead of tribal knowledge
- Plan capacity and configure autoscaling that keeps up with a real scaling event
- Run load tests whose results survive contact with production
- Inject failure deliberately with AWS FIS, and command an incident in the first fifteen minutes
- Handle gray AZ failure and make the multi-region call on evidence, then close the loop with postmortems and toil reduction
Prerequisites
- You operate something on AWS, or are about to - this course assumes a system exists
- Familiar with CloudWatch basics, load balancers and at least one compute service
- No prior SRE experience needed; the practices are built up from first principles
Course outline
Module 1: Reliability Foundations
Reliability Foundations - 2 topics.
- What SRE means on AWS specifically 8 min read
- SLIs, SLOs, and error budgets with CloudWatch 9 min read
Module 2: Instrumentation
Instrumentation - 2 topics.
- Instrumenting a distributed system - what to keep 9 min read Pro lesson
- Alerting that does not destroy your on-call rotation 8 min read Pro lesson
Module 3: Operating At Scale
Operating At Scale - 3 topics.
- Runbooks and Systems Manager Automation 9 min read Pro lesson
- Capacity planning and autoscaling that actually works 9 min read Pro lesson
- Load testing and performance baselining on AWS 8 min read Pro lesson
Module 4: Failure
Failure - 4 topics.
- Failure injection and chaos engineering with AWS FIS 8 min read Pro lesson
- Incident command on AWS - the first fifteen minutes 9 min read Pro lesson
- Multi-AZ failure modes - what actually happens 9 min read Pro lesson
- Multi-region: active-active vs active-passive 10 min read Pro lesson
Module 5: Improvement
Improvement - 2 topics.
- Postmortems and the AWS-specific failure taxonomy 9 min read Pro lesson
- Toil reduction and the automation ladder 9 min read Pro lesson