flybrain/infra/config/node-exporter-default
acamilo 660c3cf00d
Some checks failed
ci / node 22 (test + typecheck) (push) Has been cancelled
ci / rust stable (cargo test --workspace --release) (push) Has been cancelled
ci / infra/tests/lint.sh (push) Has been cancelled
ci / playwright apps/stage (allowed to fail) (push) Has been cancelled
flybrain v0.4.0: public tree (history retained privately)
2026-09-21 15:09:46 +00:00

16 lines
907 B
Text

# infra/config/node-exporter-default — pushed to
# /etc/default/prometheus-node-exporter, read by Debian's
# prometheus-node-exporter.service as $ARGS.
#
# docs/design/infra.md section 5 item 1: "prometheus-node-exporter in each fly
# CT with --collector.textfile.directory=/var/lib/node_exporter/textfile,
# listening on the CT address :9100". 02-base.sh installs the package, but
# nothing set the flag, so the textfile collector was off and the watchdog's
# fly_watchdog_*, the retention pruner's and fly-backup-stage's .prom files
# would never have been scraped even once the directory existed. Found on the
# P0 spike run 2.
#
# The listen address is left at the package default (:9100 on all
# interfaces) deliberately: the container has no public route, and the metrics container's
# Prometheus has to reach it across the LAN.
ARGS="--collector.textfile.directory=/var/lib/node_exporter/textfile"