I've worked in SRE/infra for the last few years, handled real...
Yoshik@AskYoshik
26 views
Aug 15, 2026
~1 min read
Advertisement
1
I've worked in SRE/infra for the last few years, handled real production systems, interviews, outages, migrations, all of it.
If I had to guide a fresher into DevOps in 2026 from zero, this is the exact order I'd make them follow:
If I had to guide a fresher into DevOps in 2026 from zero, this is the exact order I'd make them follow:
2
Linux first.
Not "learn 50 commands"
Understand processes, permissions, filesystems, memory, CPU, logs, ports, services.
Not "learn 50 commands"
Understand processes, permissions, filesystems, memory, CPU, logs, ports, services.
3
Networking properly.
DNS, TCP, HTTP, TLS, routing, NAT, subnets, load balancers.
I keep repeating this because most "DevOps problems" eventually become networking problems once you go deep enough.
DNS, TCP, HTTP, TLS, routing, NAT, subnets, load balancers.
I keep repeating this because most "DevOps problems" eventually become networking problems once you go deep enough.
4
Git + Bash.
You don't need to become a shell wizard.
But you should be comfortable with branches, merges, rebasing, pipes, grep, curl, awk, basic scripting.
You'll use this stuff constantly.
You don't need to become a shell wizard.
But you should be comfortable with branches, merges, rebasing, pipes, grep, curl, awk, basic scripting.
You'll use this stuff constantly.
5
Docker before Kubernetes.
Build an image.
Run it.
Break it.
Exec inside it.
Understand ports, volumes, networks, PID 1, why containers stop.
Build an image.
Run it.
Break it.
Exec inside it.
Understand ports, volumes, networks, PID 1, why containers stop.
6
Pick one cloud.
AWS, GCP or Azure, doesn't matter that much at first.
Learn compute, networking, IAM, storage, load balancers, databases.
Don't try to memorise 150 service names.
AWS, GCP or Azure, doesn't matter that much at first.
Learn compute, networking, IAM, storage, load balancers, databases.
Don't try to memorise 150 service names.
7
Learn CI/CD by shipping something.
Build → test → artifact → deploy → rollback.
One working pipeline you understand properly is worth more than watching 10 CI/CD tutorials.
Build → test → artifact → deploy → rollback.
One working pipeline you understand properly is worth more than watching 10 CI/CD tutorials.
8
Terraform/OpenTofu.
Learn state, plan, apply, modules, drift, dependencies.
The important part is not writing HCL.
Learn state, plan, apply, modules, drift, dependencies.
The important part is not writing HCL.
9
Now learn Kubernetes.
Pods, deployments, services, ingress, probes, requests/limits, ConfigMaps, secrets.
Then spend more time debugging Kubernetes than creating resources.
Pods, deployments, services, ingress, probes, requests/limits, ConfigMaps, secrets.
Then spend more time debugging Kubernetes than creating resources.
10
Learn observability + incidents.
Logs, metrics, traces, alerts.
But more importantly learn to answer:
what changed?
where is the request slow?
Logs, metrics, traces, alerts.
But more importantly learn to answer:
what changed?
where is the request slow?
11
I also share practical Linux, networking, DevOps, SRE resources, deep dives and useful engineering projects every week on Uptime Engineer.
Join 1,000+ engineers:
uptime.engineer
Join 1,000+ engineers:
uptime.engineer