Sreeram K R

Infrastructure Engineer

Right now: two free-tier hosts, one Hermes box behind no open ports, snapshots every six hours.

I build infrastructure that stays cheap and hard to reach. This site is the case study: a portfolio that moved from a hand-built VM to the edge, then back onto two free-tier hosts that cost nothing and expose nothing. Every phase is written down, including the ones I got wrong.

TerraformGCPOracle Cloud CloudflareGitHub Actions

View Resume

sreeramkr: deploy

$0per month, both hosts 6hsnapshot cadence 8phases documented 0open ports

The current fleet. One Hermes host that restores itself from a six-hourly snapshot, and the GCP box running Vaultwarden. Each is reachable only through its own outbound tunnel.

The Story

Eight phases, from a hand-deployed VM to two free-tier hosts with no open ports. Every phase has its own diagram, and the reversals are in there too.

01Bootstrap & GitOps polling Historical
A VM that redeployed itself whenever the repo changed. It worked, and I could not tell which commit was live.
02Observability & IaC serverless Historical
First I could see what the box was doing. Then I rewrote the deployment in Terraform and tried serverless.
03No-Open-Ports Docker Compose Historical
The serverless bill arrived. Everything went back to one VM, this time with no open ports at all.
04Edge migration to Cloudflare Pages Historical
The site moved to Cloudflare Pages, which freed the VM for real work.
05Repurposing the compute Historical
Two idle VMs got jobs: a Telegram assistant, then an Oracle dev box.
06Vaultwarden host Live
The GCP box became a password vault. Loopback only, with nightly off-site backups.
07Oracle dev box + Hermes Historical
A dev box with three ways in: a browser terminal, a harness, and Hermes in Docker.
08Consolidation Live
All three became one. Native Hermes, a snapshot every six hours, and a rebuild that restores itself.

Read the full archive →