Skip to content

Release Notes

v2.0.0

This is a major release that introduces a unified virtbench CLI, three new test scenarios (Chaos, FIO, Elbencho), a vm-ops day-2 operations command group, a rewritten failure-recovery test, and a full documentation site.

New Features

Unified virtbench CLI

A single Click-based entry point that wraps every benchmark and operation in the suite. Subcommands:

  • virtbench datasource-clone — VM creation / boot storm
  • virtbench migration — live migration testing
  • virtbench chaos-benchmark — concurrent chaos workload
  • virtbench failure-recovery — node-failure recovery with FAR
  • virtbench fio — FIO storage I/O benchmark
  • virtbench elbencho — Elbencho storage benchmark
  • virtbench validate-cluster — pre-flight cluster validation
  • virtbench vm-ops <op> — day-2 VM operations group
  • virtbench version — version information

Top-level options (--log-level, --log-file, --kubeconfig, --timeout, --uuid) apply across all subcommands.

The virtbench CLI is the documented entry point for user workflows. Scenario examples now use virtbench commands instead of direct Python script execution.

New: Chaos Benchmark

virtbench chaos-benchmark runs concurrent chaos operations — VM creation, volume resize, volume clone, VM restart, and snapshots — to stress-test cluster resilience under mixed concurrent load.

New: Elbencho Storage Benchmark

virtbench elbencho runs the elbencho storage micro-benchmark across multiple VMs to measure file-system level throughput and latency, paired with the existing FIO scenario.

New: vm-ops Day-2 Operations Group

A dedicated command group for VM lifecycle operations used during benchmarking and validation:

  • vm-ops drain-nodes — drain Kubernetes nodes and measure drain time
  • vm-ops rebalance-vms — rebalance VMs across worker nodes
  • vm-ops vm-snapshot — create VirtualMachineSnapshots in batches
  • vm-ops run-blkdiscard — run blkdiscard on data disks inside VMs
  • vm-ops power-toggle-vms — power VMs on or off

Each operation is exposed through the virtbench vm-ops command group.

vm-ops rebalance-vms now targets worker nodes by default, computes an automatic target range that handles small VM counts, and only includes master/control-plane nodes when --include-master-nodes is provided.

New: Boot Storm Testing Scenario

Boot storm functionality (--boot-storm on datasource-clone) is now documented as a first-class scenario with single-node, multi-node, and existing-VM variants. See the dedicated Boot Storm guide.

Creation and boot-storm result paths now use the storage driver label and disk count instead of the storage class name:

results/{storage-driver}/{num-disks}-disk/{timestamp}_{namespace-prefix}_{start}-{end}/

Run logs are saved in the same timestamped run folder as the JSON and CSV result files.

Live Migration Updates

virtbench migration now exposes the migration modes needed for common operator workflows, including auto-select-busiest node evacuation, round-robin migration, interleaved scheduling, and multi-source-node migration.

Multi-source-node migration accepts a comma-separated --source-nodes value, for example:

virtbench migration --source-nodes worker-1,worker-2,worker-3

It discovers matching VMIs directly from the cluster, removes VM/VMI nodeSelector settings before migration, and submits migrations in an interleaved order across source nodes.

The migration guide now includes an end-to-end workflow covering cluster validation, VM creation, boot storm, rebalance, and multi-source-node live migration against the same VM set.

Results and Logging

Saved benchmark runs now keep logs, detailed results, and summary results together inside the same run directory. Commands executed through virtbench are logged with sensitive values redacted where applicable.

The public result grouping option is standardized as --storage-driver across documented workflows. The previous --storage-version spelling is removed from the user documentation.

Rewritten: Failure Recovery

failure-recovery/recovery-test.py replaces the previous measure-recovery-time.py and the run-far-test.sh / run-manual-failure-test.sh / patch-vms.sh shell scripts. The wrapper exposes node-targeted recovery testing via virtbench failure-recovery, including --mode monitor, --mode manual, and --mode far-operator.

Documentation Site

Full MkDocs Material site under docs/, including:

  • Repository Structure reference
  • Test Scenarios index with per-scenario guides
  • VM Operations sub-section
  • Configuration Options reference
  • Output and Results, Cleanup Guide, VM Template Guide
  • Best Practices and Troubleshooting

Structural Changes

  • I/O benchmarks consolidated under io-benchmark/{fio,elbencho}/. The legacy fio-benchmark/measure-fio-performance.py location is removed.
  • Failure-recovery shell scripts removed in favour of a single Python test driver.
  • New top-level vm-ops/ directory containing the day-2 operation scripts.
  • Hotplug disk support removed from vm-ops/ and from documentation.

Breaking Changes

  • fio-benchmark/measure-fio-performance.py → io-benchmark/fio/measure-fio-performance.py
  • failure-recovery/measure-recovery-time.py, run-far-test.sh, run-manual-failure-test.sh, patch-vms.sh → replaced by failure-recovery/recovery-test.py
  • --storage-version documentation replaced with --storage-driver.
  • Some default values changed for datasource-clone and migration — see Configuration Options for the per-flag defaults of both the wrapper and the underlying scripts.

Notes

  • The virtbench CLI is the recommended and documented entry point for benchmark workflows. User-facing examples use virtbench commands.