Get in touch →
The physical server running Albgott — a Dell PowerEdge T160

ENGINEERED TO RUN.
DESIGNED TO LAST.

For many years, we’ve been building and operating Albgott, a private cloud platform — improving it step by step, learning from what works, and making it better with every iteration.

See the architectureRead the log
Services in production
38
Network segments
08
Operated since
2023
Monitored
24/7

ALBGOTT is a private cloud built from the ground up and operated in-house.

It started as infrastructure for running a few services.Three years later, it has grown into a complete platform — with isolated networks, dedicated compute, its own access model, and observability across every layer.

It has been running in production since 2023, and it keeps evolving one improvement at a time.

  • Compute03 NODES
  • Network08 SEGMENTS
  • Workloads38 SERVICES
  • ObservabilityFULL STACK · 24/7
  • AutomationIN PROGRESS
01 — Pillars

The principles behind it

01

Architecture

Isolation by design. Networks, systems, and services have their own boundaries. Nothing shares a layer without a reason.

Explore →02

Security

Trust isn't implicit. Being connected does not mean being allowed. Every connection is explicitly defined, controlled, and limited to what is actually needed.

Explore →03

Operations

Built to keep running. Systems are continuously monitored, failures are surfaced early, and the infrastructure keeps getting better over time.

Explore →
02 — SystemFull breakdown →

System topology

Compute
pve01pve02 pve03
Network
Publicingress
Internalapps
Datastateful
CIbuild
Backuprestricted
Corpwork
Mgmtadmin
Homeuser
Access
AdministratorsDevelopers

Deliberate
trade-offs

Isolation over convenience. Some decisions make things harder to operate. They also make mistakes smaller, access more controlled, and failures easier to contain.

  • 01
    Asymmetric node roles
    Compute goes where the workload belongs — not simply where it is most convenient.
  • 02
    No implicit trust
    Two networks do not communicate just because both are considered “internal.” Access is explicit by design.
  • 03
    Access split by role
    Administrators and developers use separate access paths, keeping credentials, permissions, and blast radius apart.
03 — SecurityFull breakdown →

Nothing gets in by accident.

access-policy.conf
# default posture
deny   *              → *

# explicit exceptions only
allow  public→ app :443
allow  internal→ tools, trusted net only
allow  corp-tunnel→ internal
allow  admin-tunnel→ mgmt console

# identity
rollout sso, secrets→ in progress, not finished

How trust
is controlled

Being on the network is not the same as being allowed to use it. Every boundary below exists on purpose.

  • 01
    Default deny, explicit allow
    Every connection between segments exists because it was written down, not because it was convenient.
  • 02
    Trust tiers, not one router doing everything
    Public traffic, internal tools, and hardware management sit behind three separate entry points, each with its own blast radius.
  • 03
    Identity, still rolling out
    Single sign-on and centralized secrets exist and work today — adopting them everywhere is honest, ongoing work, not a finished migration.
04 — OperationsSee operations →

OBSERVE.AUTOMATE.ALERT.

Every layer is monitored, measured, and instrumented to surface problems early before it becomes operational debt.

PROMETHEUSGRAFANALOKIJENKINSUPTIME KUMAALERTMANAGER
PROMETHEUSGRAFANALOKIJENKINSUPTIME KUMAALERTMANAGER
05 — LogView all →

Latest from the log

Postmortem2026.08.19

Disaster Recovery Before Real Backups

Runbooks vs. guides, a general failure-triage flow, and why documentation-as-source-of-truth is a compensating control, not a backup substitute.

Read →
Decisions2026.08.12

Client Isolation Tiers

What changes when a mostly-personal platform starts hosting a workload for an external party, and why onboarding and offboarding are mirrors of each other.

Read →
Operations2026.08.05

Alert Fatigue by Design

Using inhibition rules so a critical alert silences its own warning-level echo instead of paging on-call twice for one root cause.

Read →
06 — RoadmapFull roadmap →

Not finished. Said out loud.

20/ 30+ goals shipped

There's no backup system running today. If a node goes down, whatever lived only there is gone — not degraded, gone. That gap is public on purpose, and so is everything else still open.

  • ✓Split the network into isolated zones
  • ✓Build a complete CI/CD pipeline
  • ✓Watch the whole stack — metrics and alerts
  • Turn SSO on for every service
  • Get backups running automatically
  • Make the monitoring stack survive a node dying
  • Move the hardware into a proper rack
07 — AboutThe rest of the story →

Built in the open, mistakes included.

Every outage, wrong call, and unfinished piece gets written down where it happened — not smoothed over afterward.

— A.M.

Built by Alberto Morán — Software Engineer

GOT SOMETHING THAT NEEDS TO RUN?

If you're working on infrastructure, platforms, or services where reliability matters beyond day one, we'd like to hear what you're building.

Get in touch →