From 69c00616b0ed2cf867e9cd1b602999ff2236182c Mon Sep 17 00:00:00 2001 From: clover caruso Date: Tue, 20 Oct 2026 14:27:24 -0700 Subject: [PATCH] Record VM recovery and Windows preset rollout evidence Record production console continuity through a real route reload and the prepared Windows template validation. Preserve the distinction between template timings and a controlled SSD-cache benchmark. Assisted-by: gpt-6-astra --- tools/mac-agent-vm.md | 8 +++++++ tools/windows-presets.md | 50 ++++++++++++++++++++++++++++++++++++++++ tools/zfs-read-cache.md | 4 ++++ 3 files changed, 62 insertions(+) create mode 100644 tools/windows-presets.md diff --git a/tools/mac-agent-vm.md b/tools/mac-agent-vm.md index 6494d741ea1365aa1114e15b0900c48d106002c6..96fd8437faed52567e8afe68bcbfcebb0f63817c 100644 --- a/tools/mac-agent-vm.md +++ b/tools/mac-agent-vm.md @@ -93,6 +93,14 @@ Power operations are serialized; a timeout kills unfinished lifecycle commands while preserving the detached VM runner. Hardware edits, media attachment, force-off, deletion, and preset creation are unavailable for this managed VM. +Dashboard WebSockets survive Caddy route reloads for up to their existing +eight-hour session limit. Interrupted viewers reconnect with bounded backoff +and recheck authorization. On 2026-10-20, removing the completed dashboard +preview reloaded production routes while the Mac viewer continued the same +video stream without reconnecting. A deliberately closed viewer connection +also recovered without restarting the guest. The idle legacy tailnet relay +was retired; the actively used loopback recovery console was preserved. + ## Guest access `~/.local/bin/code-vm` links to the installed `runtime/control/code-vm` under diff --git a/tools/windows-presets.md b/tools/windows-presets.md new file mode 100644 index 0000000000000000000000000000000000000000..a3ed4d97630bf3e565ec215d28bb2fc87334d929 --- /dev/null +++ b/tools/windows-presets.md @@ -0,0 +1,50 @@ +# Windows presets + +Prepared images live in the VM library's `Presets` directory. Each published +directory is immutable: existing guests retain QCOW2 backing-file references +to it. Publish a new ID with `supersedes` to change the library default while +preserving older bases. Host paths come from the configured VM library root. + +The provisioning contract is [windows-provision.ps1](../guest/windows-provision.ps1). +Prepare an isolated template with its account named `sandbox`, then run that +script with `-Prepare` from elevated 64-bit Windows PowerShell. Preparation +randomizes the template password, removes clone identity and completion state, +and installs the startup task. Shut down normally before capturing through +`vms.save_preset` with `setup=windows-v3`. The dashboard deliberately rejects +ordinary Windows preset capture because it cannot establish preparation. + +Validate fresh clones with different account names before publication. Check +the personalized account and computer name, the `ready` marker, Explorer, +the guest capture process, and the actual dashboard screen and input. Compare +the published disk and NVRAM bytes with the tested candidate, and check that +the disk is standalone with `qemu-img info` and `qemu-img check`. Never replace +an existing base in place. + +## 2026-10-20 rollout + +The `windows-10-pro-v10` candidate was prepared and tested on HDD storage with +four vCPUs, 4 GiB RAM, and a 64 GiB disk. Both its base and each clone's writable +overlay stayed on HDD. The host's 150 GiB SSD read cache remained enabled. + +Preparation refreshed the 64-bit .NET Framework native-image cache with +`ngen update /force` and installed native images for the assemblies loaded by +Windows PowerShell, then reran the normal `-Prepare` flow. All assembly-install +commands and preparation completed successfully. See Microsoft's +[Ngen documentation](https://learn.microsoft.com/en-us/dotnet/framework/tools/ngen-exe-native-image-generator) +for regeneration after framework servicing. No global CLR switches were added. + +| Candidate | Fresh clone to ready | Account | +| --- | ---: | --- | +| v9 repeat | 455 s | fixture | +| v10 first | 206 s | fixture | +| v10 repeat | 161 s | fixture2 | + +Both v10 clones completed personalization and reboot, started Explorer and the +guest capture process, and received distinct computer names. These are rollout +observations with differing cache warmth, not a controlled measurement of +L2ARC's contribution. Synchronous writes still reach HDDs. + +The published v10 preset supersedes v4. v4 and earlier referenced bases remain +available to their existing guests. Reverting the default requires withdrawing +v10 from discovery while preserving its bytes wherever clones still reference +them; it does not require modifying existing guest disks. diff --git a/tools/zfs-read-cache.md b/tools/zfs-read-cache.md index b206f5f456271103df3f5a6382c124a19a95f79a..3754c2ee7d9bf965ca25f76dcf17f136697ce0e5 100644 --- a/tools/zfs-read-cache.md +++ b/tools/zfs-read-cache.md @@ -80,3 +80,7 @@ the production browser console, accepted a keyboard command, and shut down normally. Its definition and disk were removed afterward. The release's service health checks passed. Cache warming was observed; these checks do not establish a Windows startup speedup. + +Subsequent template preparation and fresh-clone timings are recorded in +[Windows presets](windows-presets.md). Those observations do not isolate the +read cache's contribution. -- 2.54.1