After the power cut, power is back, every machine is on… and the laboratory software is stuck on “502”

Annour Biologie groups six medical testing laboratories in Fez and Meknes. All results go through the LIS (laboratory information system): https://sil.annour.ma, served by rp-01 (nginx) in front of two application servers app-sil-01/app-sil-02, a PostgreSQL database on db-01 (data on an iSCSI volume provided by stor-01), the PDF reports on an NFS share of stor-01, and the directory + internal DNS annour.lan on dc-01. All these machines are VMs on the Proxmox hypervisor hv-01, in the technical platform's server room.

Last night, mains power cut from 02:50 to 05:10 across the neighbourhood. The UPS lasted 40 minutes, the generator did not start (broken since August). At 05:10, power comes back: hv-01 boots and starts all the VMs at the same time. This morning, every machine answers ping, but the LIS shows “502 Bad Gateway”. The first samples arrive at 07:00, and the emergency department of the nearby hospital is waiting for test results.

You are the on-call systems administrator on poste-astreinte, with root SSH everywhere. Bring the service back cleanly, layer by layer, without putting the database's data at risk.