How do I avoid losing files if a VM or node fails?

To avoid losing files when a VM or node fails, you want to make storage independent of any single machine and ensure it’s replicated/backed up.

Best practices

1. Don’t store important data only on local disk

If a VM’s disk lives only on that VM’s host, a host failure can take the data with it. Use:

• Shared/network storage: NFS, SMB, iSCSI, SAN
• Distributed storage: Ceph, GlusterFS, vSAN, Longhorn, Portworx, etc.
• Cloud managed disks/object storage: EBS + snapshots, Azure Disks + backups, GCP Persistent Disks + snapshots, S3/Blob/GCS for files

2. Use replication

Keep multiple copies of the data on different physical nodes or availability zones.

• For databases/filesystems, use built-in replication if available
• For VM platforms, use HA/clustering plus replicated storage
• Aim for at least 3 copies for critical data where possible

3. Back up regularly

Replication helps with hardware failure, but not with:

• accidental deletion
• corruption
• ransomware
• bad deployments

So also do:

• scheduled backups
• offsite backups
• versioned backups / snapshots
• immutable backups if possible

A good rule is: 3-2-1

• 3 copies of data
• 2 different media/storage types
• 1 copy offsite

4. Use snapshots carefully

Snapshots are useful for quick rollback, but they are not a substitute for backups.

• Good for short-term recovery
• Not ideal as the only protection
• Make sure snapshots are stored on separate resilient storage

5. For VMs, enable HA / failover

If the node running a VM dies:

• use a hypervisor cluster with high availability
• ensure the VM can restart on another node
• store the VM disk on shared or replicated storage

Examples:

• VMware HA + vSAN/NFS/SAN
• Proxmox HA + Ceph
• Hyper-V cluster + SMB3/S2D
• KVM clusters with Ceph or shared storage

6. For apps, design for statelessness

If possible:

• keep app servers stateless
• store user data in databases or object storage
• put uploaded files in replicated storage, not local VM disks

7. Test restores

A backup is only useful if you can restore it.

• periodically test recovery
• verify backup integrity
• document restore procedures

Simple rule of thumb

If losing one VM or node must not lose files:

• store data outside that node
• replicate it
• back it up offsite
• test recovery

If you want, I can suggest the best setup based on your environment, like:

• VMware / Proxmox / Hyper-V
• Kubernetes
• cloud (AWS/Azure/GCP)
• bare metal or NAS/SAN