# Notes: Home dashboard consumes `infra_*` entities for exceptions-only alerts.
# Notes: Home dashboard consumes `infra_*` entities for exceptions-only alerts.
# Notes: WAN outages alert immediately at near-zero throughput or 90% packet loss; performance degradation requires 6 continuous hours below 300/300 Mbps, above 80 ms, above 5% loss, or unavailable telemetry.
# Notes: WAN outages require all four independent ICMP/HTTPS probes to fail continuously for 2 minutes.
# Notes: Continuous latency/loss determine WAN quality; Ookla throughput is a separate twice-daily capacity snapshot and never declares an outage.
# Notes: Severe quality turns critical after 30 minutes above 250 ms or 25% loss; moderate quality degradation requires 6 hours above 80 ms or 5% loss.
# Notes: Nightly Duplicati verification runs at 08:00 after the 05:30 Duplicati job and docker_14 reboot window.
# Notes: Nightly Duplicati verification runs at 08:00 after the 05:30 Duplicati job and docker_14 reboot window.
# Notes: Duplicati transport/API errors are logged only; repairs are reserved for proven failed or stale backups.
# Notes: Duplicati transport/API errors are logged only; repairs are reserved for proven failed or stale backups.
# Notes: Duplicati failure Repairs enable a recovery poll that clears the Repair after a later successful run.
# Notes: Duplicati failure Repairs enable a recovery poll that clears the Repair after a later successful run.
@ -82,6 +84,28 @@ command_line:
state_class:measurement
state_class:measurement
value_template:"{{ (value | regex_replace('[^0-9.]', '')) or 'unknown' }}"
value_template:"{{ (value | regex_replace('[^0-9.]', '')) or 'unknown' }}"
kernel_update_packages={{ kernel_update_packages | trim }}
kernel_update_packages={{ kernel_update_packages | trim }}
required_policy=This dispatch is Carlo's authorization to install pending same-release Proxmox package updates after live preflight passes; do not stop after read-only validation. Inspect live Proxmox cluster health, storage, VM placement, and node status before changing anything. Patch hosts one node at a time during the overnight window and verify each node remains healthy before moving to the next. Do not perform major-version upgrades, repository migrations, destructive cleanup, or force operations from this request. If the live preflight is unhealthy, pause and report instead of patching. If kernel packages are present or a reboot is required after kernel updates, use the $kernel-refresh skill and follow its ordered node-cycling workflow. Verify the live placement of Frigate/Docker14 VM 101, stop it before rebooting the host that owns it, and start it only after the host cycle and VM migrations are complete. After each host, verify the package versions/update list, then report progress, failures, resume state, package versions, and final verification.
required_policy=This dispatch is Carlo's authorization to install pending same-release Proxmox package updates after live preflight passes; do not stop after read-only validation. Inspect live Proxmox cluster health, storage, VM placement, node status, and destination memory headroom before changing anything. Patch hosts one node at a time during the overnight window and verify each node remains healthy before moving to the next. Do not perform major-version upgrades, repository migrations, destructive cleanup, or force operations from this request. If the live preflight is unhealthy, pause and report instead of patching. If kernel packages are present or a reboot is required after kernel updates, use the $kernel-refresh skill and follow its ordered node-cycling workflow; if that skill is unavailable, stop before any reboot and report the blocker.
migration_policy=Capture each running VM's expected final node before evacuation. Migrate exactly one VM at a time and wait for task success plus running/unlocked/HA-started state before starting another. If any return migration fails or any non-Docker14 VM remains on ProxMox02, recover the expected placement and recheck memory headroom before continuing; do not start Docker14 while placement is incomplete.
docker14_policy=Verify the live placement of Frigate/Docker14 VM 101, gracefully shut it down before rebooting either Proxmox node even when VM 101 is hosted on the other node, confirm Proxmox reports it stopped, and start it only after the final host cycle, sequential return migrations, and capacity checks are complete. Starting the existing stopped VM 101 for this authorized recovery does not require separate confirmation. Never issue a Proxmox reboot while Docker14 is running.
completion_policy=After each host, verify package versions and the update list. Final success requires expected VM placement, no locks, HA resources started, Docker14 and dependent services healthy, and enough host memory headroom; report progress, failures, resume state, package versions, and final verification.
Carlo Costanzo
2 months ago