VMware backup, snapshot and restore planning for STACKIT

What the run knows about snapshots, and what it does not

An eight-machine RVTools export written for this page and assessed by StackToIT on 4 November 2026, with the sheets vSnapshot and vMemory included. Left: the vSnapshot sheet. Right: the snapshot findings recorded in assessment.json (finding identifiers left out).

VM,Name,Description,Date / time,Filename,Size MiB (vmsn),Size MiB (total),Quiesced,State,VM UUID
db-pg01,before-upgrade-15,Before PostgreSQL 15 minor upgrade,2026-06-02 21:14:00,[ds-core-01] db-pg01/db-pg01-Snapshot1.vmsn,33000,48200,False,poweredOn,8004-eeee-0001
db-pg01,pre-reindex,Before reindex,2026-09-18 22:03:00,[ds-core-01] db-pg01/db-pg01-Snapshot2.vmsn,33000,131400,True,poweredOn,8004-eeee-0001
file01,Backup-test,Restore test,2026-10-30 08:20:00,[ds-core-01] file01/file01-Snapshot1.vmsn,8200,9100,True,poweredOn,8004-eeee-0005
test-clone07,VM Snapshot 3/12/2026,,2026-03-12 11:00:00,[ds-core-01] test-clone07/test-clone07-Snapshot1.vmsn,4100,22500,False,poweredOff,8004-eeee-0008

[
  {
    "id": "vm.db-pg01",
    "findings": [
      {
        "severity": "info",
        "message": "The VM has snapshots; consolidate them before export — only the current state is migrated."
      }
    ]
  },
  {
    "id": "vm.file01",
    "findings": [
      {
        "severity": "info",
        "message": "The VM has snapshots; consolidate them before export — only the current state is migrated."
      }
    ]
  },
  {
    "id": "vm.test-clone07",
    "findings": [
      {
        "severity": "info",
        "message": "The VM has snapshots; consolidate them before export — only the current state is migrated."
      }
    ]
  }
]

• Count per machine

  The run found 4 snapshots on 3 of the 8 machines and raised one note on each of the three: consolidate before export, only the current state is migrated.

• Age, size and quiescing are not in the note

  The note is the same for a snapshot of 5 days and one of 237 days. The sheet holds the date, the total size and whether it was quiesced. Read it yourself.

• Against the vendor's advice

  Broadcom advises keeping a single snapshot no longer than 72 hours. On 4 November 2026 the four snapshots are 155, 47, 5 and 237 days old, and the largest of them reports 131400 MiB in total.

Backup scope is the set of volumes in the draft

From the same run. Left: the two disks of db-pg01. Right: the server and the volume the engine drafted for it. The draft holds no backup or snapshot resource.

VM,Disk,Capacity MiB,Raw,Disk Mode,Sharing mode,Thin,Controller,Shared Bus,VM UUID
db-pg01,Hard disk 1,102400,False,persistent,sharingNone,True,SCSI controller 0,noSharing,8004-eeee-0001
db-pg01,Hard disk 2,614400,False,persistent,sharingNone,True,SCSI controller 1,noSharing,8004-eeee-0001

resource "stackit_server" "vm_db_pg01" {
  project_id        = var.project_id
  name              = "db-pg01"
  machine_type      = var.vm_db_pg01_machine_type
  availability_zone = var.availability_zone
  boot_volume = {
    source_type = "image"
    source_id   = var.vm_db_pg01_image_id
    size        = 100
  }
  network_interfaces = [stackit_network_interface.vm_db_pg01_nic1.network_interface_id]
}

resource "stackit_volume" "vm_db_pg01_disk2" {
  project_id        = var.project_id
  name              = "db-pg01-disk2"
  availability_zone = var.availability_zone
  size              = 600
  performance_class = var.vmware_volume_performance_class
}

• Two volumes to protect

  The boot volume of 100 GiB and the data volume of 600 GiB. STACKIT backs up per volume, so both belong in the schedule.

• Backups are yours to add

  Nothing in the four Terraform files defines a backup. The STACKIT page on Server Backup Management says the service is also available in the CLI and the Terraform provider.

• One performance class

  The volume reads var.vmware_volume_performance_class, shared by all data volumes. STACKIT states that backup and restore time depends on volume size and performance class.

Four protections that sound alike

Statements about STACKIT are as of November 2026 and follow the Server introduction, the Server Backup Management pages and the Server FAQ listed under sources. The Server introduction also notes that the backups and snapshots feature is being integrated into the Portal, and it states a maximum of 6 snapshots per volume; the IaaS API how-to shows the API route. Check limits again before you rely on them.

1 · Discovery

• Both sheets are in the export

  vSnapshot and vMemory. Without them the run warns that snapshots and memory reservation of every VM are unknown, not none. Look for the warning in engineering/architect-report.md.

• Every machine with a snapshot is listed

  The snapshot note in assessment.json names the units. Here: db-pg01 with two, file01 and test-clone07 with one each. Each gets an owner and a yes or no.

• Age, size and quiesce flag are read from the sheet

  RVTools_tabvSnapshot.csv holds date, sizes and quiesce state. The run counts the snapshots and adds up their sizes in the unit data of assessment.json (db-pg01: 2 snapshots, 179600 MiB), but its notes do not mention them.

• Today's protection is written down per machine

  Tool, schedule, retention and last successful restore test. RVTools does not hold this. The run takes RPO and tolerated downtime only from an applications file you provide.

2 · Design

• Each snapshot use has a successor

  Pre-change rollback, test copies, backup by habit. Decide for each: STACKIT volume snapshot, volume backup or nothing.

• The backup tool is chosen

  Server Backup Management, your existing backup software or database-native backup. STACKIT says databases on a server should be backed up separately.

• Schedule and retention per tier

  Write them against the RPO of each application. Server Backup Management allows custom schedules and a freely chosen retention.

• Scope matches the volumes in the draft

  Compare terraform/compute.tf with your backup scope: every boot volume and every data volume, for db-pg01 two.

• A restore time target exists

  STACKIT states that restore duration depends on volume size, volume performance type and load of the backup system. Size the performance class with that in mind.

3 · Build

• Backup definitions live next to the infrastructure

  The draft contains none. Add them to your own Terraform or scripts and review them like any other change.

• Volume performance class and sizes are confirmed

  The class is one variable, the sizes come from the source capacity rounded up. Both appear as settings left to you.

• The snapshot decision is recorded

  Each machine with snapshots, here db-pg01, file01 and test-clone07, has an open record in review/decisions.yaml that lists the snapshots setting. Fill resolution and rationale, then run again.

4 · Data

• Snapshots are consolidated before export

  Only the current state migrates. Broadcom warns against cancelling a consolidation and notes that snapshots made through backup software may not show in Snapshot Manager, so check from the command line too.

• Databases have an application-level copy

  For db-pg01, a dump or the database's own backup, next to the volume copy. A crash-consistent volume backup is a fallback, not the plan.

• A last source backup exists and is kept

  Take it before the final copy and keep it until a restore on STACKIT has been proven.

5 · Cutover

• A quiet window is agreed where it matters

  STACKIT recommends the server with the volume be switched off before a volume backup, otherwise the result is crash-consistent. Decide per application.

• The first STACKIT backup runs right after cutover

  Check that its status reaches available, as the IaaS API how-to describes.

• The source machine and its backups stay

  Keep both until the new backups cover the period you need. Retire them by decision, not by habit.

6 · After

• A restore has been performed and timed

  Create a new volume from a snapshot or backup, attach it to a server and check the data. The IaaS API how-to lists the steps.

• Retention behaves as designed

  Backups are deleted when their retention ends. Confirm that the oldest kept copy is the one you planned.

• Old snapshots, backups and runbooks are retired

  Remove VMware snapshots and old backup jobs by decision and update the restore runbook to the STACKIT steps.

Questions about backups and snapshots

Sources

• STACKIT docs: Introduction to Server (snapshots and backups)
• STACKIT docs: Server Backup Management
• STACKIT docs: Server FAQ
• STACKIT docs: Volume snapshots and backups via the IaaS API
• Broadcom: Best practices for using VMware snapshots