authoravatarclover caruso<git@paperclover.net>2026-10-20 14:27:24-07:00
committeravatarclover caruso<git@paperclover.net>2026-10-21 02:52:29-07:00
log69c00616b0ed2cf867e9cd1b602999ff2236182c
treed6a09e02c8eacf3f8a7a5851859837b2b1a4ffe2
parentd5255060f16bc0ddac53d35a07477df8c5d131e3
signature Signed by SSH key SHA256:52mNGHRsVFBDED9IAX5pe+LRWUefqTbxEReunq21QvU

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

3 files changed, 62 insertions(+), 0 deletions(-)

tools/mac-agent-vm.md+8
......@@ -93,6 +93,14 @@ Power operations are serialized; a timeout kills unfinished lifecycle commands
9393while preserving the detached VM runner. Hardware edits, media attachment,
9494force-off, deletion, and preset creation are unavailable for this managed VM.
9595
96Dashboard WebSockets survive Caddy route reloads for up to their existing
97eight-hour session limit. Interrupted viewers reconnect with bounded backoff
98and recheck authorization. On 2026-10-20, removing the completed dashboard
99preview reloaded production routes while the Mac viewer continued the same
100video stream without reconnecting. A deliberately closed viewer connection
101also recovered without restarting the guest. The idle legacy tailnet relay
102was retired; the actively used loopback recovery console was preserved.
103
96104## Guest access
97105
98106`~/.local/bin/code-vm` links to the installed `runtime/control/code-vm` under
tools/windows-presets.md created+50
......@@ -0,0 +1,50 @@
1# Windows presets
2
3Prepared images live in the VM library's `Presets` directory. Each published
4directory is immutable: existing guests retain QCOW2 backing-file references
5to it. Publish a new ID with `supersedes` to change the library default while
6preserving older bases. Host paths come from the configured VM library root.
7
8The provisioning contract is [windows-provision.ps1](../guest/windows-provision.ps1).
9Prepare an isolated template with its account named `sandbox`, then run that
10script with `-Prepare` from elevated 64-bit Windows PowerShell. Preparation
11randomizes the template password, removes clone identity and completion state,
12and installs the startup task. Shut down normally before capturing through
13`vms.save_preset` with `setup=windows-v3`. The dashboard deliberately rejects
14ordinary Windows preset capture because it cannot establish preparation.
15
16Validate fresh clones with different account names before publication. Check
17the personalized account and computer name, the `ready` marker, Explorer,
18the guest capture process, and the actual dashboard screen and input. Compare
19the published disk and NVRAM bytes with the tested candidate, and check that
20the disk is standalone with `qemu-img info` and `qemu-img check`. Never replace
21an existing base in place.
22
23## 2026-10-20 rollout
24
25The `windows-10-pro-v10` candidate was prepared and tested on HDD storage with
26four vCPUs, 4 GiB RAM, and a 64 GiB disk. Both its base and each clone's writable
27overlay stayed on HDD. The host's 150 GiB SSD read cache remained enabled.
28
29Preparation refreshed the 64-bit .NET Framework native-image cache with
30`ngen update /force` and installed native images for the assemblies loaded by
31Windows PowerShell, then reran the normal `-Prepare` flow. All assembly-install
32commands and preparation completed successfully. See Microsoft's
33[Ngen documentation](https://learn.microsoft.com/en-us/dotnet/framework/tools/ngen-exe-native-image-generator)
34for regeneration after framework servicing. No global CLR switches were added.
35
36| Candidate | Fresh clone to ready | Account |
37| --- | ---: | --- |
38| v9 repeat | 455 s | fixture |
39| v10 first | 206 s | fixture |
40| v10 repeat | 161 s | fixture2 |
41
42Both v10 clones completed personalization and reboot, started Explorer and the
43guest capture process, and received distinct computer names. These are rollout
44observations with differing cache warmth, not a controlled measurement of
45L2ARC's contribution. Synchronous writes still reach HDDs.
46
47The published v10 preset supersedes v4. v4 and earlier referenced bases remain
48available to their existing guests. Reverting the default requires withdrawing
49v10 from discovery while preserving its bytes wherever clones still reference
50them; it does not require modifying existing guest disks.
tools/zfs-read-cache.md+4
......@@ -80,3 +80,7 @@ the production browser console, accepted a keyboard command, and shut down
8080normally. Its definition and disk were removed afterward. The release's
8181service health checks passed. Cache warming was observed; these checks do not
8282establish a Windows startup speedup.
83
84Subsequent template preparation and fresh-clone timings are recorded in
85[Windows presets](windows-presets.md). Those observations do not isolate the
86read cache's contribution.