Skip to content

Compute disposition review (owner decisions)

Status: Owner decisions recorded (2026-06-23) Prepared: 2026-06-23 Execution queue: compute decommission queue Sources: prox live scan, whrrr VMM scan, infra-services guest facts, lab audit, inventory YAML

Use this document to record keep / consolidate / retire decisions per compute instance. After you respond, we update inventory status, runbooks, and proxmox consolidation execution order.


How to respond

For each row, set Owner decision to one of:

Code Meaning
K Keep as-is (instance stays on current HV)
K-s Keep but leave stopped until needed
C→ Consolidate — move workload to target in Consolidate target column
R Retire — destroy instance after backup/verification window
D Defer — need more discovery (in-guest facts, owner interview)

Reply in chat or edit this file inline, e.g. k6-loadtest: R or fill the Owner decision column below.

Normal vs notable services: Normal = expected baseline for that instance type. Notable = duplicates, experiments, unknown purpose, or non-standard deployments worth a disposition call.

Patch status legend:

Label Meaning
Wave N In Phase 8 coordinated patching (patching_waveN)
Unmanaged Not in ansible patch orchestration — patch age unknown
Appliance Vendor/OS-managed updates only (HAOS, DSM)
Retired Instance gone or decommissioned

Summary matrix

Prox resource pressure (from lab audit): ~95% RAM, ~89% local-lvm — stopping or retiring idle guests frees allocatable vCPU/RAM/disk even when powered off.

HV Host VMID / guest Type Power OS (inventory) Ansible Patch status Age / provenance Notable services Agent rec. Consolidate target Owner decision
prox saltierpoop 100 VM running Ubuntu 22.04 managed (OS) Wave 0 Pre-2026 Saltbox Saltbox media stack (~40 containers) — normal K K
prox infra-services 123 VM running Ubuntu 24.04 managed Wave 2 (C&C) Provisioned Phase 4 traefik, komodo, ara, homepage, adguard, monitoring (13 containers) — normal K K
prox haos 200 VM running HAOS appliance Appliance Pre-2026 Home Assistant — normal K K
prox influxdb 111 LXC running Debian unmanaged Unmanaged Pre-2026 InfluxDB (2 orgs: hypervisor-monitoring, nfs-monitoring); Telegraf writers D D — keep until dashboard migration (see below)
prox harbor-registry 119 LXC running Ubuntu unmanaged Unmanaged Pre-2026 reg.realemail.app — recordurbate-tiktok depends K K
prox pulse 116 LXC running Debian unmanaged Unmanaged Pre-2026 Unknown — high network I/O; role undocumented D K
prox k6-loadtest 105 LXC running Ubuntu unmanaged Unmanaged Pre-2026 k6 load testing (lab-audit: decommission) R — (or ad-hoc binary on workstation) R - but backup to somewhere
prox nfs-monitoring 114 LXC stopped Ubuntu unmanaged Unmanaged Pre-2026 NFS bridge to Whrrr (synorpn + prawns) K-s K-s
prox octoprint 120 LXC stopped Debian unmanaged Unmanaged Pre-2026 OctoPrint + USB passthrough for 3D printer K-s K-s
prox metrimon 106 VM stopped Ubuntu 24.04 unmanaged Unmanaged Pre-2026 Legacy monitoring VM R infra-services (Prometheus/Grafana already live) R - but everything must migrate
prox dnsproject 107 VM stopped unknown unmanaged Unmanaged Community script DNS side project (pre-AdGuard) D AdGuard on infra-services OR retire R - but backup somewhere
prox penpot 121 VM stopped unknown unmanaged Unmanaged Community script Penpot design tool C→ infra-services Docker stack R - but backup somewhere
prox aiproject 102 LXC stopped Ubuntu unmanaged Unmanaged Pre-2026 AI project workspace K-s Merge with ollama later (Phase 9) R
prox ollama 115 LXC stopped Ubuntu unmanaged Unmanaged Pre-2026 Ollama LLM K-s Single AI VM with aiproject (Phase 9) K-s
prox unmanic 103 LXC stopped Debian unmanaged Unmanaged Pre-2026 Media transcoder C→ saltierpoop Docker (Tdarr already there) R
prox graylog 109 LXC stopped Debian unmanaged Unmanaged Experiment Central syslog (not SIEM) K Revive LXC 109 K — Pattern E: Graylog syslog + Wazuh SIEM
prox caddy 117 LXC stopped Debian unmanaged Unmanaged Experiment Reverse proxy R infra-services Traefik R - but backup somewhere
prox mysql 113 LXC stopped Debian unmanaged Unmanaged Experiment Standalone MySQL R saltierpoop MariaDB or infra-services if needed K-s
prox sqlserver2022 110 LXC stopped Ubuntu unmanaged Unmanaged Experiment SQL Server (23 GB RAM alloc) R — (recreate on demand if ever needed) K-s
prox reactive-resume 118 LXC stopped Debian unmanaged Unmanaged Experiment Resume builder app C→ infra-services compose OR retire (SaaS) R - but backup somewhere
prox netboot.xyz 122 LXC stopped Debian unmanaged Unmanaged Experiment PXE/iPXE boot server R infra-services on-demand container R
prox blocktopus 104 LXC destroyed Debian n/a Retired PiHole → AdGuard cutover Jun 2026 R — (already destroyed) R
prox prox bare metal running Debian 12 managed Wave 1 Hardware 2024+ Proxmox VE hypervisor — normal K K
whrrr recordurbate VMM VM running Ubuntu unmanaged Unmanaged Customer app Host for recordurbate-tiktok metrics stack K K, might be worth migrating to Prox, though. Prox may have better resource capacity
whrrr ubuncap VMM VM running Ubuntu unmanaged Unmanaged Customer app Docker: recordurbate-tiktok compose (/mnt/streams/tiktok/) K K, s/a
whrrr whrrr NAS/appliance running DSM 7.3.2 managed (OS) Wave — RS2421+ VMM, Plex, Docker, volumes — normal K K

Detail cards (expand for context)

Production — no change expected

saltierpoop (VM 100)

Field Value
Resources 20 vCPU, 30 GB RAM (balloon), 260 GB disk, GPU passthrough
Normal services Saltbox: traefik, authentik, plex, *arr stack, mariadb, etc.
Notable Duplicate legacy prometheus/grafana/loki containers (Saltbox-managed) — awareness only
Patch Wave 0; OS-layer ansible only — do not touch /opt/*
Recommendation Keep — production media stack

infra-services (VM 123)

Field Value
Resources 4 vCPU, 8 GB RAM, 30 GB disk
Normal services traefik, komodo (+mongo, periphery), ara, homepage, adguard (+unbound), prometheus, grafana, loki, promtail, alertmanager
Patch Wave 2; last coordinated patch pass live 2026-06-23
Guest facts infra-services.json (2026-06-23)
Recommendation Keep — canonical infra + monitoring C&C

haos (VM 200)

Field Value
Resources 6 GB RAM, 32 GB disk; USB passthrough (Zigbee)
Normal services Home Assistant
Patch Appliance — manual HAOS updates
Recommendation Keep — home automation critical path

Running — needs owner input

influxdb (LXC 111) — running

Field Value
Resources 2 GB RAM, 8 GB disk
Notable services InfluxDB; buckets hypervisor-monitoring, nfs-monitoring; Grafana datasources on infra-services still reference it
Open question Are Telegraf agents / dashboards still actively using this vs Prometheus on infra-services?
Recommendation Defer — run in-guest discovery + check Grafana datasource usage before retire vs keep

pulse (LXC 116) — running

Field Value
Resources 1 GB RAM, 4 GB disk; IP 192.168.6.199
Notable services Unknown — lab audit noted heavy network I/O; inventory role still unknown
Discovery gap patch-controller SSH to root not authorized — in-guest facts not collected
Recommendation Defer — identify workload, then keep vs retire

k6-loadtest (LXC 105) — running ⚠️

Field Value
Resources 6 GB RAM, 12 GB disk
Notable services k6 load generator
Conflict Lab audit marked decommission; instance is powered on as of 2026-06-23 scan
Recommendation Retire unless you still run load tests — k6 is a single binary; no dedicated LXC needed

harbor-registry (LXC 119) — running

Field Value
Resources 4 GB RAM, 80 GB disk; IP 192.168.6.119
Notable services Harbor at reg.realemail.app
Dependency recordurbate-tiktok images
Recommendation Keep until registry moves to infra-services or external

Stopped — consolidation candidates

Host VMID RAM Disk Agent rec. Consolidate target Rationale
graylog 109 8 GB 30 GB Retire Loki on infra-services Superseded by Phase 5 stack
caddy 117 2 GB 12 GB Retire Traefik on infra-services Superseded
metrimon 106 10 GB 96 GB Retired infra-services monitoring Destroyed 2026-06-24; duplicate stack
mysql 113 1 GB 8 GB Retire saltierpoop MariaDB Duplicate DB
sqlserver2022 110 23 GB 60 GB Retire Largest idle alloc; no known consumers
reactive-resume 118 3 GB 8 GB Retired Owner 2026-06-24: destroy without backup (SaaS if needed)
penpot 121 4 GB 10 GB Retired Owner 2026-06-24: destroy without backup
netboot.xyz 122 0.5 GB 8 GB Retire On-demand container Rarely needed
unmanic 103 4 GB 8 GB Consolidate saltierpoop Docker Same media domain as Tdarr
dnsproject 107 3 GB 20 GB Retired Owner 2026-06-24: vzdump then destroy (pre-AdGuard experiment)
aiproject 102 9 GB 64 GB Keep stopped Future merged AI VM Phase 9 RAG
ollama 115 10 GB 35 GB Keep stopped Future merged AI VM Phase 9

nfs-monitoring (LXC 114) — stopped

Field Value
Notable Active NFS bridge when running — synorpn + prawns mounts to Whrrr
Recommendation Keep stopped until next monitoring/NFS initiative; do not retire without mount migration plan

octoprint (LXC 120) — stopped

Field Value
Notable Privileged LXC for USB serial to printer
Recommendation Keep stopped — power on when printing; not a consolidation candidate

Whrrr VMM

recordurbate / ubuncap

Field recordurbate ubuncap
IP 192.168.6.98 192.168.6.100
VMM guest_id 2c23e005-2de9-49d2-9872-995e6892209e a5dc6c98-5628-4aa9-9601-a668f73983ce
Workload Customer-app host recordurbate-tiktok Docker compose
Patch Unmanaged Unmanaged
Recommendation Keep both — customer-app boundary; not prox consolidation scope

Projected savings (if agent recommendations accepted)

Rough allocated resources recovered if all Retire rows are destroyed and consolidation targets absorb the Consolidate rows:

Category Instances ~RAM ~Disk
Retire (stopped experiments) graylog, caddy, metrimon, mysql, sqlserver2022, netboot.xyz, reactive-resume† ~47 GB ~224 GB
Retire (running surprise) k6-loadtest ~6 GB ~12 GB
Consolidate then retire shell unmanic, penpot, reactive-resume† ~11 GB ~26 GB

† reactive-resume counted once in retire-after-consolidate path.

End-state sketch (Phase 9+): prox runs ~8–10 instances: saltierpoop, infra-services, haos, harbor-registry, influxdb (if kept), pulse (if kept), nfs-monitoring, octoprint, aiproject+ollama (one future AI VM), plus hypervisor.


Discovery gaps (fill before deciding)

Host Gap Command / action
pulse, influxdb, harbor-registry, … No in-guest JSON yet Authorize patch-controller key on root; discover-guests.yml
influxdb Unknown if still referenced Check Grafana datasources + Telegraf configs on prox
dnsproject Retired 2026-06-24 vzdump on prox before destroy; purpose never documented
k6-loadtest Running vs decommission intent Your call
saltierpoop containers Not enumerated in repo Optional: one-shot docker ps artifact (Phase 3.3 plan)

Owner decisions (2026-06-23)

Recorded from matrix Owner decision column. Inventory updated; destroy work tracked in compute decommission queue.

Host Decision Notes
saltierpoop, infra-services, haos, prox, whrrr K Production
harbor-registry, pulse K pulse: document workload via discover-guests
nfs-monitoring, octoprint, ollama, mysql, sqlserver2022 K-s Stopped; keep allocated
recordurbate, ubuncap K Optional Prox migration Phase 9+
k6-loadtest, dnsproject, penpot, aiproject, unmanic, caddy, reactive-resume, netboot.xyz R Backup before destroy where noted
metrimon R Migrate all workloads first
graylog K Central syslog on LXC 109; Wazuh for SIEM
influxdb D → keep for now Cannot retire until Grafana dashboards migrate
blocktopus R Already destroyed

Open items (agent responses)

influxdb — “alternative target?”

You cannot retire Influx yet. Grafana on infra-services still queries LXC 111 today:

  • Datasource influxdb-nfs-monitoring → org nfs-monitoring (NFS Monitoring dashboard)
  • Datasource proxbox-influxdb → org hypervisor-monitoring, bucket thermal-data (Proxbox Thermals dashboard)

Alternative path (recommended order):

  1. Keep influxdb running until dashboards are rewritten.
  2. Migrate writers — point Telegraf agents (prox thermals, NFS probes) at Prometheus remote write or direct scrape on infra-services.
  3. Rewrite dashboards — Proxbox Thermals + NFS Monitoring to Prometheus datasources.
  4. Then retire influxdb LXC and remove Grafana Influx datasources.

Prometheus is already the canonical metrics store for Phase 5; Influx is a legacy dependency for two dashboards, not a parallel platform you must replace with another TSDB.

graylog + SIEM — resolved (Pattern E, 2026-06-23)

Decision: Both — Graylog stays central syslog (revive LXC 109); Wazuh is SIEM on managed Linux. Loki unchanged for container ops.

Layer Tool Runbook
Central syslog Graylog LXC 109 central-syslog-graylog.md
SIEM Wazuh on infra-services wazuh-siem.md

Inventory: graylog active, role central-syslog. Removed from decommission queue.


Changelog

Date Change
2026-06-23 Owner decisions recorded; inventory status/lifecycle updated

Superseded paste block

# Decisions captured in table + Owner decisions section above.