DevOps / SRE / Platform Engineer
A connected route from foundations to production systems: 13 stages and about 248 hours, including practice and projects. Learn, practise, and prove each stage.
Stage 1 of 13
Set a target on the planner
Draft. This path has not been reviewed by a second person yet. Topic order, estimates and tasks may change.
The list view has the same topics and works with any screen reader.
Stage 1: Operating a host
3 topics · 18 h · milestone: Run a service in WSL under systemd0 / 3- Not started2 h
- Not started17 min
- Not started12 min
Stage project and rubric
Run a service in WSL under systemd. Restart it, diagnose a permissions fault and leave a tested recovery script.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The service starts at boot, restarts after you kill it, and its journal shows both, on a clean WSL instance.
- You diagnose the permissions fault from process, file and user evidence, and the fix grants no more than the service needs.
- The recovery script is idempotent: running it twice leaves the host in the same working state.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 2: The network path
4 topics · 18 h · milestone: Connect two local services through a reverse proxy0 / 4- Not started58 min
- Not started17 min
- Not started36 min
- Not started20 min
Stage project and rubric
Connect two local services through a reverse proxy. Diagnose a DNS, TLS or firewall misconfiguration from observations.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- Both services answer through the proxy, and a capture or log shows each hop the request takes.
- For the injected fault you name the layer from observations, before changing anything.
- You write the checklist you followed, in order, so someone else could repeat the diagnosis.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 3: Shared changes
2 topics · 8 h · milestone: Land a reviewed configuration change in a disposable repository and recover its previous behaviour0 / 2- Not started3 h
- Not started34 min
Stage project and rubric
Land a reviewed configuration change in a disposable repository and recover its previous behaviour.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The change lands through a reviewed branch, with a note saying what changes, how to verify it and how to undo it.
- You resolve a real conflict deliberately and explain why the kept lines are right.
- You restore the previous behaviour with history intact, and show the commit that did it.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 4: The service you operate
3 topics · 14 h · milestone: Build and run the small typed service the later stages package, deploy and observe, and explain each part of a request through it0 / 3- Not started3 h
- Not started14 min
- Not started21 min
Stage project and rubric
Build and run the small typed service the later stages package, deploy and observe, and explain each part of a request through it.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The service builds and runs from a clean checkout with one documented command.
- You trace one request from socket to handler to response, naming each layer.
- A typed input error is rejected with a clear status and message, and a test proves it.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 5: Packaging services
3 topics · 18 h · milestone: Start an app and database with Compose0 / 3- Not started55 min
- Not started52 min
- Not started11 min
Stage project and rubric
Start an app and database with Compose. Preserve a record after replacing the application container.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The image runs as a non-root user, reports healthy, and stays under its size budget; the verifier passes.
- A record survives replacing the database container, and the verifier's --after run passes.
- You list the image's inputs: pinned base digest, SBOM packages, and what a signature would prove.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 6: Delivery
2 topics · 16 h · milestone: Release two versions to local slots0 / 2- Not started22 min
- Not started20 min
Stage project and rubric
Release two versions to local slots. Shift traffic, observe health and roll back.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The same immutable build is promoted from test to the live slot; nothing is rebuilt on the way.
- You shift a share of traffic, watch a stated signal, and roll back within a time you measured.
- You compare rolling, blue-green, canary and GitOps for this service and justify your choice.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 7: Cloud fundamentals on AWS
4 topics · 26 h · milestone: Design a private service and a least-privilege role0 / 4- Not started2 h 07
- Not started18 min
- Not started39 min
- Not started13 min
Stage project and rubric
Design a private service and a least-privilege role. Verify the policy against real requests and estimate the cost locally.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The diagram shows every route, subnet and gateway, and a request's path through them is labelled.
- The role's policy passes the IAM practice checker's least-privilege tests for your service.
- You estimate the monthly cost from stated inputs and name the largest driver.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 8: Infrastructure changes
2 topics · 20 h · milestone: Provision local resources with OpenTofu0 / 2- Not started52 min
- Not started15 min
Stage project and rubric
Provision local resources with OpenTofu. Detect drift and converge twice without recreating data.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- A second apply reports no changes; nothing holding data is destroyed or recreated.
- You introduce drift by hand, show it in a plan, and converge it back.
- State is kept outside the working directory, and you explain how you would recover it.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 9: Operating a cluster
3 topics · 36 h · milestone: Deploy an app in kind with a volume0 / 3- Not started40 min
- Not started1 h 54
- Not started47 min
Stage project and rubric
Deploy an app in kind with a volume. Recover a failed rollout and verify service access.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The app reaches Ready behind a Service, and survives deleting one pod with its data intact.
- You diagnose and fix a failed rollout from events and probes, then roll back with Helm.
- A team role can deploy but cannot read secrets, proven by a refused request.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 10: Signals in production
2 topics · 18 h · milestone: Build a dashboard and alert for a failing worker0 / 2- Not started17 min
- Not started1 h 04
Stage project and rubric
Build a dashboard and alert for a failing worker. Link one sampled request from metric to trace.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The dashboard shows the worker's rate, errors and queue depth, from scraped metrics.
- The alert fires when the worker stops, and its for duration is justified against noise.
- You follow one slow request from a metric to its trace and name the slow span.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 11: Reliability practice
3 topics · 24 h · milestone: Lead a local incident drill and write a blameless report with a measurable toil reduction0 / 3- Not started21 min
- Not started19 min
- Not started51 min
Stage project and rubric
Lead a local incident drill and write a blameless report with a measurable toil reduction.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The objective is stated as an SLI, a target and a window, with the error budget it leaves.
- The drill has a timeline, a named incident commander and a restore within the target you set.
- The postmortem is blameless and ends with a toil-reducing action and how it will be measured.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 12: Secure operations
3 topics · 16 h · milestone: Deploy a least-privilege workload0 / 3- Not started3 min
- Not started3 min
- Not started25 min
Stage project and rubric
Deploy a least-privilege workload. Verify its policy and replace a test secret without exposing it in logs.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- The workload runs under the restricted pod security level, and a privileged pod is refused.
- A test secret is rotated, and you show it never appears in logs, images or the repository.
- A policy-as-code rule refuses an unpinned image, with a test for both outcomes.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Stage 13: A platform for other engineers
2 topics · 16 h · milestone: Deliver a service template with a pipeline, dashboards and an owner0 / 2- Not started15 min
- Not started8 min
Stage project and rubric
Deliver a service template with a pipeline, dashboards and an owner. Have another service adopt it locally.
Use your own small service or a disposable local example. Score each criterion: 0 missing, 1 partial, 2 demonstrated.
- A second service adopts the template and reaches a healthy deployment; you measure how long it took.
- The template has a named owner, a support route and a documented upgrade path.
- You decide whether to add a service mesh, with the evidence and the signal that would change your mind.
Target: 2 on every criterion. Record your evidence in the notes of the stage's last topic. Report a problem with this project.
- Not started
Source inventory and coverage
36 topics map to 93 unique activities. Shared resources count once. Every topic has a self-check; every stage has a project rubric.
Supplementary pages kept outside this path:
Pages under Library and Practice come from, or build on, Donne Martin's System Design Primer (CC BY 4.0).