We are running Postgres 15.4 with a timescale extension on an Ubuntu server (22.04). At random intervals the database slows down: running a select 1; query from a local client may take up to 3-4 minutes. Database returns to normal on its own after a short period.

All OS vitals reported by Zabbix at that moment seem normal (CPU load, available memory, disk usage). Number of connections to the database is at 20% of the maximum. During that period however new connections to the database time out and zabbix stops reporting vitals from the database itself. Does anybody have any idea of other metrics I should be looking into?