←→ move · scroll · M map · N notes · F fullscreen

All slides

Speaker notes

24 October 2026 · server.sentrico.ai · 45.63.31.253

Server
Report.

Everything this box serves, examined one vhost at a time — 54 vhosts surveyed, 53 sites photographed, every cert, DNS record and verdict.

16keep
10review
24moved off-box
3broken / retired

01 · The host

A modest box with
a serious job list.

4vCPU EPYC-Milan
7.2 GBRAM · 4.8 used + 3.3 swap
60%root disk · 97/169 GB
51 duptime · Ubuntu 26.04.1

Web tier

nginx fronts 54 vhosts · MariaDB · PHP 8.5-fpm · TLS via certbot + Cloudflare origin certs

App tier

22 Docker containers across 9 compose stacks · systemd apps (Bun site, Retention Bee) · vite/Next dev servers

Guard rails

fail2ban · Tailscale · Authentik SSO · acme.sh + certbot renewals

02 · Identity

Two jobs,
one box.

Production web host

StaffConnect HRMS & portals, Sentrico properties, client microsites — real client traffic lands here.

AI agent workstation

Hermes (:9119) · Honcho memory + Ollama embeddings · Neo4j · headless Chrome fleet · CLI proxy — and the RAM bill that comes with it.

03 · Method

How the survey ran.

Probe

Every vhost checked for HTTP status, page title, TLS cert expiry, and public DNS — 53 domains, 3 data tables

Photograph

53 screenshots captured against local nginx by four parallel subagents, two re-shoot passes for expired-cert and origin-cert vhosts

Verdict

Each site graded KEEP · REVIEW · MOVED · BROKEN/RETIRED with the evidence attached

Screenshots show what this server serves — where public DNS points elsewhere, that difference is itself evidence.

04 · The keep list

16 sites earn
their keep.

5StaffConnect client stack
4core sentrico.ai properties
5AI infrastructure
2WFR apps

These carry real traffic: the Frappe HRMS with 9 containers, Laravel portals, the Hermes agent platform, Neo4j Bolt, and the WFR technician system — all healthy at survey time.

Keep · Client platforms

StaffConnect runs here.

Keep · Properties & AI infrastructure

The brand, and the
machinery behind it.

Keep · Applications

Live applications,
not demos.

24

05 · The big finding

MemberPulse
has moved.

All 24 memberpulse.com.au domains — portals, admin, WordPress — resolve away from this server. What remains here are stale copies with certificates that died in April 2026.

Moved · Evidence

DNS already told
the story.

memberpulse.com.aupublic DNS
→
149.28.176.224portals ×20 · wordpress ×4
Verceladmin.memberpulse

Certs expired

Every local memberpulse cert lapsed 11–12 April 2026 — renewal broke the moment DNS moved, because validation followed the traffic

Zero traffic here

Nothing public reaches these vhosts — they serve stale copies of portals that live elsewhere now

Moved · The estate

All 24, still parked here.

admin
client
app · 502
members.wt
aala admin
aala member
aala sponsor
aala client
aala wp
cba admin
da member
aslm sponsor

4 orgs (aala · aslm · cba · da) × 4 portals + WordPress, plus 4 root domains — retire the vhosts and the memberpulse-platform copy once the new box is confirmed.

410

06 · Retired, but still running

SuiteCRM said
goodbye. Its containers
didn't.

The vhost serves 410 Gone, DNS is deleted — yet 4 containers + MariaDB still run 24/7 for it. Decommissioning them is the cheapest win on this box.

Repair · 502s

Two dead ends.

graph.sentrico.ai

Neo4j Browser proxy → 502 · Neo4j not running · restart it or retire the vhost

buzz.sentrico.ai

502 · upstream down · identify purpose or remove

07 · Review pile — 10 sites

Keep-or-kill calls.

Also: sopos-map (auth wall) · decks (403) · 3rdarmor (Cal.com embed) · rb (Retention Bee prototype)

08 · Housekeeping

Small leaks, taken
together.

Memory pressure

3.3 GB of swap in active use — Chrome fleet, Ollama and 22 containers compete for 7.2 GB

Dead mount

~/mnt/laptop sshfs mount dropped — every df errors with "Transport endpoint is not connected"

Zombie cargo

7 exited alpine containers (9–10 months old) · pm2 daemon running with zero apps · 45-day-old headless Chrome fleet

Cert watch

Renewals work where DNS still points here — memberpulse's April expiry is what moving-away looks like

09 · The plan

What to do next.

1 · Decommission

SuiteCRM stack (4 containers + MariaDB) · 24 memberpulse vhosts + platform copy · 7 dead alpine containers

2 · Repair

graph (restart Neo4j or drop) · buzz (fix or remove) · clear the dead sshfs mount · retire idle pm2

3 · Consolidate

workingtitle → one domain · archive nexted-v1 · decide the review pile (omac, zenplanner, 3rdarmor, decks, sopos-map)

4 · Watch

RAM + swap after decommissioning · cert renewals · disk at 60% and climbing

End of deck

The receipts,
one click away.

Every screenshot, cert status, DNS note and verdict — filterable — lives at
server-report.sentrico.ai

Surveyed & generated automatically · 24 October 2026 · server.sentrico.ai