US Sr SRE (devops) checking in: I would personally recommend the networking path. Caveat: A good engineer will know the background of both (curl, telnet, Iam, security groups, cidrs, domains)
Devops was mostly automating the stuff in between the other teams; and most of that is working out of the box these days. Most repos already have their Jenkins and docker files. How much admin are you expecting on serverless? Most people are pivoting to app support (ticket queues) or supporting managed services (on call).
As far as my day to day:
- Troubleshooting incidents and walk ups
- Answering pages (read restarting things)
- Groovy Jenkins build pipelines
- Cdk applications
- Ruby configuration management
- Parameter/secret management
- Reading error messages for devs
- Yaml/xml linting
- Assisting in load testing
- Changing settings to make the application more stable. Ex: db connections, memory
- Cloud UI/clis
Pros: I do a lot of different things, we get downtime because we need to respond to things immediately, I don't have normal project/sprint planning. I have the keys to the kingdom. Higher pay than most other devs. I hack things together, I don't need to design workflows.
Cons: I am on call, I am the silliest clown (I get hardest problems), I need to understand a lot of moving pieces, sometimes when things break, there is a lot of pressure on you to find something hard. I regular have to Google "bash variable syntax" because I'm coding in 15 languages. Interviewing for jobs is impossible because no 2 positions are the same
I personally don't like the idea of migrating off Jenkins, we blew our yearly budget testing our build platform in git. But it's all just platformed ci/CD, which is why I'm recommending the other path. Platform teams lost the goal recently.