Monitoring, availability and backups
Keep the grid accessible and recover when a service or disk fails.
Monitor what actually matters
Track CPU, RAM, swap, disk space, inodes, storage latency and networking on the physical host and any VMs. A process can be running while its SQL connection or region access is broken. Combine metrics with application-level tests and logs.
Uptime and downtime
A personal PC available only evenings has different expectations from a public 24/7 grid. Set reasonable availability goals or published hours and announce maintenance. Renting from a provider does not alone eliminate outages.
Back up more than just VMs
Back up ROBUST SQL databases, FSAssets payloads and metadata, INI configs, service data and possibly OAR/IAR archives. A VM snapshot or quick backup may not be application-consistent across SQL and FSAssets while writes continue without coordination. Test restoration in isolation.
Protect against complete host loss
Keep at least one backup independent from the primary host. If VMs and all backups live on the same physical server or disk, a hardware failure can take everything down. Encrypt sensitive backup data and manage retention, access and verification.
Non-destructive verification commands
These Linux commands provide an initial non-destructive picture. Compare readings to the normal baseline; low free RAM is not automatically trouble because Linux uses memory for cache.
Illustrative example: adapt to the actual installation.
uptime free -h vmstat 1 5 df -h df -i ip -brief address ss -lnt # Adapter le service pour lire ses journaux : journalctl -u NOM_SERVICE -n 60 --no-pager
Recovery plan and service order
Document who can intervene, where backups are kept, emergency credentials and recovery commands. Check storage and SQL health before restarting writers. Effective recovery follows the actual dependencies between ROBUST, the database and simulator processes rather than blind restarts.