For a week, every ERP night job has failed: the VM is switched off every morning

The finance department of MasterLinux runs its ERP on the Ubuntu VM ml-erp-01 (group rg-erp-prod, region francecentral). Every evening at 22:00, the systemd timer ml-batch-nuit.timer starts the daily closing: invoicing of enrolments, accounting export, reports for management.

The VM was created on Monday 12 October by the infrastructure team to replace the old server. Since the next day, no night job has completed: every morning, the scheduler ml-ordo-01 reports “ml-erp-01 unreachable”, and the on-call person restarts the VM from the portal. Everything works during the day… then it starts again in the evening. This morning, the accountant raises the alarm: the weekly closing does not exist and the VAT return must be sent on Friday.

The ticket is yours. You are on your workstation with the Azure CLI signed in to the masterlinux-prod subscription.