Oskar Kapala
e5ecefe2b7
Observer now optionally (PROM_SHADOW_URL) queries Prometheus up{} once per
cycle and LOGS SHADOW_LIVENESS_MISMATCH when its event-driven liveness
disagrees. Parallel-run only: compute_liveness and _emit_node_transition are
untouched; authoritative liveness stays 100% event-driven. Fail-open on any
Prometheus error (down/timeout/bad JSON -> {}). 9 new tests, incl. proof that
shadow-read does not change node_info liveness/status.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
47 lines
1.8 KiB
YAML
47 lines
1.8 KiB
YAML
# Control-plane production overrides for the VPS deployment.
|
|
#
|
|
# NODE_ALIAS_MAP translates the node names that appear in raw event files
|
|
# (written by node agents / seed scripts) to the canonical names used in
|
|
# inventory/topology.yaml and hosts/*/services.yaml.
|
|
#
|
|
# Current live mapping (from /opt/homelab/events/ inspection):
|
|
# node-2 → chelsty (zigbee2mqtt / mosquitto / homeassistant node)
|
|
#
|
|
# Add further entries when new nodes come online and their event-source names
|
|
# differ from their topology names. Format is a single-line JSON object, e.g.:
|
|
# NODE_ALIAS_MAP='{"node-2":"chelsty","node-3":"piha"}'
|
|
#
|
|
# The executor inherits the canonical name from the action JSON written by the
|
|
# supervisor, so NODE_ALIAS_MAP is only required on the supervisor service.
|
|
#
|
|
# Memory limits: VPS has 4 GiB RAM, no swap. oom_score_adj -900 ensures the
|
|
# host kernel OOM-killer never targets control-plane containers. mem_limit
|
|
# provides a per-container cgroup ceiling so a leaking process is restarted by
|
|
# Docker before it can exhaust host memory.
|
|
|
|
services:
|
|
operator-ui:
|
|
mem_limit: 192m
|
|
oom_score_adj: -900
|
|
|
|
observer:
|
|
mem_limit: 192m
|
|
oom_score_adj: -900
|
|
# SHADOW-READ (Prometheus liveness cutover — etap 1). Optional: when set, the
|
|
# observer ALSO queries Prometheus up{} each cycle and LOGS mismatches vs its
|
|
# event-driven liveness (grep SHADOW_LIVENESS_MISMATCH), but NEVER switches
|
|
# source. Unset → disabled, observer unchanged (fail-open on any prom error).
|
|
# Uncomment to enable the ≥7-day parallel-run:
|
|
# environment:
|
|
# - PROM_SHADOW_URL=http://100.95.58.48:9090
|
|
|
|
supervisor:
|
|
mem_limit: 400m
|
|
oom_score_adj: -900
|
|
environment:
|
|
- NODE_ALIAS_MAP={"node-2":"chelsty"}
|
|
|
|
executor:
|
|
mem_limit: 64m
|
|
oom_score_adj: -900
|