My 3-node Talos homelab had no monitoring, so I only found out something was wrong when an app stopped working.
So I added VictoriaMetrics for metrics and VictoriaLogs for logs from every node and pod. Both show up in Grafana, and alerts go to my phone on Telegram.
Healthchecks[.]io also watches the alerting itself, so I hear about it if the alerts stop. The whole stack uses about 1.3 GiB of RAM across the three nodes.