Run blkdiscard

Run blkdiscard against data disks inside KubeVirt VMs to release blocks back to thin-provisioned storage.

Use Case: Reclaim space after destructive write tests (FIO, elbencho), verify TRIM/UNMAP plumbing through the storage stack, or normalize a test fleet before a fresh run.

How It Works

For each VM in the namespace range the operation:

1. Connects to the VM through an SSH bastion pod (--ssh-pod in --ssh-pod-ns).
2. Auto-detects data disks (every block device except the OS disk) — or uses the explicit list from --disks.
3. Runs blkdiscard /dev/<disk> on each target disk.
4. Logs success / failure per disk.

Basic Usage

Using virtbench CLI

# Auto-detect data disks across 10 VMs
virtbench vm-ops run-blkdiscard \
  --namespace-prefix rhel-eb-filler \
  --start 1 --end 10 \
  --vm-name rhel-elbencho-1

# Target specific disks
virtbench vm-ops run-blkdiscard \
  --namespace-prefix rhel-eb-filler \
  --start 1 --end 10 \
  --vm-name rhel-elbencho-1 \
  --disks vdb --disks vdc

# Higher concurrency
virtbench vm-ops run-blkdiscard \
  --namespace-prefix rhel-eb-filler \
  --start 1 --end 50 \
  --vm-name rhel-elbencho-1 \
  --concurrency 25

Full Example with All Options

virtbench vm-ops run-blkdiscard \
  --namespace-prefix rhel-eb-filler \
  --start 1 \
  --end 50 \
  --vm-name rhel-elbencho-1 \
  --disks vdb --disks vdc --disks vdd \
  --ssh-pod ssh-test-pod \
  --ssh-pod-ns default \
  --vm-user root \
  --vm-password Password1 \
  --concurrency 25 \
  --log-file blkdiscard-2026-05-08.log

Options

Option	Description
--namespace-prefix (required)	Namespace prefix (e.g., rhel-eb-filler).
--start (required)	Start namespace index.
--end (required)	End namespace index.
--vm-name (required)	VM name in each namespace.
--disks (repeatable)	Specific disks (e.g., vdb, vdc). Auto-detects all data disks if omitted.
--ssh-pod	SSH bastion pod name (default: ssh-test-pod).
--ssh-pod-ns	SSH pod namespace (default: default).
--vm-user	VM SSH user (default: root).
--vm-password	VM SSH password (default: Password1).
--concurrency	Max concurrent operations (default: 10).
--dry-run	Show what would be done without doing it.
--log-file	Path to a log file (auto-generated if omitted).

Prerequisites

• An SSH bastion pod is running in the cluster and reachable from the control-plane host. The pod must have ssh and sshpass available.
• The target VMs accept the --vm-user / --vm-password credentials.
• Each target disk is not mounted (or blkdiscard will refuse). Stop or unmount the filesystem before running this operation.

Notes

• blkdiscard is destructive — it issues TRIM/UNMAP, which on most storage backends marks blocks as discarded. Do not run this against disks that hold data you want to keep.
• When omitting --disks, the operation skips the OS disk (typically vda).
• For large fleets, increase --concurrency cautiously — each parallel run opens a new SSH connection through the bastion pod.

See Also

• Cleanup Guide