First round of monitoring follow-ups from the #2 discussion (thanks @NotRetarded):
- Monitor alerts now carry the last 10 log lines — crash, exit, OOM and unhealthy notifications attach the same log tail update failures already do. The first question after a crash is always why.
- OOM alerts name the culprit, not just the victim. When one container eats the host memory and a different one gets OOM-killed for it, the alert now takes a single
docker statssnapshot at event time and lists the top memory consumers:Top memory at event time: sencho 5.1GiB · unifi 1.9GiB. No idle polling — one call when it fires. /eventscommand in Telegram — parity with the Web UI's Container Events history.