PLAN — Decommission Jamie & Foreman

I. Forensic Assessment

Jamie (jamie.1.foo)

Machine: n2-standard-2, us-central1-a, 20GB disk
Cost: ~$50/month (2 vCPUs, the most expensive non-GPU machine in the fleet)
IP: 35.193.160.134 (static: jamie-ip)
Contents:

events/ — 16,317 relay event files (3.4GB). Complete group chat archive Feb 25 – present. Actively receiving new events.
group-attachments/ — 311 media files (389MB). Group chat images/videos.
No services, no bots, no custom software, no crontab.

Purpose: Secondary relay events mirror. Same data that exists on Walter (~/events/) and vault (/mnt/events-relay/).
Verdict: Redundant. The events exist in at least two other locations. Jamie is a $50/month backup of data that's already backed up.

Foreman (foreman.1.foo)

Machine: e2-small, me-west1-b, 20GB disk
Cost: ~$15/month (1 vCPU)
IP: 34.165.28.230 (static: foreman-ip)
Contents:

Incus (container manager) installed via startup script — never configured beyond initial install
SSH key authorized but currently not working (broken for weeks, on TODO list)
Web interface at https://foreman.1.foo — basic auth (daniel/letsroll2026, chris/shitcoinporn123)
No containers running. No workloads. No data.

Purpose: Chris from SCP set this up as a Docker/Incus-based fleet management tool. Never got past initial install. Never understood how to use it. Never used it.
Verdict: Empty. A startup script ran apt-get install incus and that's the entire history of this machine.

II. What We Get Back

3 vCPUs freed — Jamie (2) + Foreman (1). Global quota goes from 8/12 (67%) to 5/12 (42%).
~$65/month saved — Jamie (~$50) + Foreman (~$15).
2 static IPs freed — jamie-ip + foreman-ip.
40GB disk freed — 20GB each.
Headroom for Ghost Jr. — with 7 free vCPUs, we can comfortably run Ghost Jr. as e2-small (1 vCPU) with room to spare.

III. Execution Plan — Jamie

STEP 1 — VERIFY EVENTS EXIST ELSEWHERE

Before touching Jamie, confirm the events archive exists on at least two other machines:

Walter: ls ~/events/ | wc -l — should be ~16,000+
Vault: ls /mnt/events-relay/ | wc -l — should be ~16,000+

If either copy is missing or incomplete, rsync from Jamie first.

STOP — VERIFY BEFORE PROCEEDING

Why: The prime directive. Jamie has 16K irreplaceable event files. We do not delete anything until we've confirmed redundancy.

What happens: Daniel reviews the verification output. Confirms events exist in 2+ other locations.

How to continue: Daniel says "verified, proceed."

STEP 2 — SNAPSHOT JAMIE'S DISK

Create a final snapshot before deletion. Belt and suspenders.

gcloud compute disks snapshot jamie --zone=us-central1-a --snapshot-names=jamie-final-20260317

Snapshots are cheap (~$0.026/GB/month = ~$0.52/month for 20GB). This is insurance.

STEP 3 — STOP THE RELAY TO JAMIE

Whatever is writing events to Jamie needs to be updated to stop. Check vault's relay service configuration and remove Jamie as a destination.

STEP 4 — STOP THE VM

gcloud compute instances stop jamie --zone=us-central1-a

Stop first, don't delete. Let it sit stopped for 48 hours. If nothing breaks, proceed to deletion.

STOP — 48-HOUR SOAK

Why: If anything depends on Jamie that we didn't notice, it'll become apparent within 48 hours.

What happens: Jamie sits stopped. Monitor for errors in relay, events, or anything else.

How to continue: 48 hours pass with no issues → proceed to delete.

STEP 5 — DELETE VM + RELEASE IP

Delete the instance (disk goes with it — snapshot is our backup):

gcloud compute instances delete jamie --zone=us-central1-a

Release the static IP:

gcloud compute addresses delete jamie-ip --region=us-central1

STEP 6 — CLEAN UP DNS + GOLD FILE

Remove jamie.1.foo DNS record from Cloudflare. Remove Jamie from fleet-gold.json. Update clankers.discount dashboard.

IV. Execution Plan — Foreman

STEP 1 — VERIFY NOTHING ON DISK

SSH is broken, so use serial console or just check from GCP metadata. The startup script only installed Incus. No data was ever created. But verify anyway.

STEP 2 — SNAPSHOT (OPTIONAL)

Foreman has nothing on it. A snapshot is probably unnecessary but costs $0.52/month and follows the prime directive. Your call.

STEP 3 — DELETE VM + RELEASE IP

No soak period needed — Foreman hasn't done anything since creation. Nothing depends on it.

gcloud compute instances delete foreman --zone=me-west1-b

gcloud compute addresses delete foreman-ip --region=me-west1

STEP 4 — CLEAN UP DNS + GOLD FILE + PASSWORDS

Remove foreman.1.foo and subdomains from Cloudflare. Remove from fleet-gold.json. The credentials (daniel/letsroll2026, chris/shitcoinporn123) can be retired — they were only used for Foreman's web UI.

V. Risk Assessment

RISK — JAMIE EVENTS

Jamie's events folder is actively receiving new files. Something on vault is writing to it. If we delete Jamie without updating the relay config, the relay will throw errors trying to write to a dead host. This is Step 3 — stop the relay before stopping the VM.

RISK — FOREMAN WEB ACCESS

Daniel shared Foreman credentials with Chris (chris/shitcoinporn123 via /chris/ path). If Chris is still using Foreman for anything, deleting it will break his access. Likelihood: near zero — the machine does nothing.

Overall risk: LOW. Jamie is a redundant mirror. Foreman is an empty shell. Neither runs any services that other machines depend on. The snapshot-before-delete protocol ensures recoverability.