| ... | @@ -48,3 +48,37 @@ The published v10 preset supersedes v4. v4 and earlier referenced bases remain | ... | @@ -48,3 +48,37 @@ The published v10 preset supersedes v4. v4 and earlier referenced bases remain |
| 48 | available to their existing guests. Reverting the default requires withdrawing | 48 | available to their existing guests. Reverting the default requires withdrawing |
| 49 | v10 from discovery while preserving its bytes wherever clones still reference | 49 | v10 from discovery while preserving its bytes wherever clones still reference |
| 50 | them; it does not require modifying existing guest disks. | 50 | them; it does not require modifying existing guest disks. |
| | 51 | |
| | 52 | ## First-run timing after fleet retirement |
| | 53 | |
| | 54 | After [retiring the unused guests](vm-retirement.md), another disposable v10 |
| | 55 | clone reached `ready` in 95.75 seconds with the same 4-vCPU/4-GiB configuration. |
| | 56 | The host observed these phases by polling the provisioning markers: |
| | 57 | |
| | 58 | | Phase | Duration | |
| | 59 | | --- | ---: | |
| | 60 | | Create disk clone and start | 1.12 s | |
| | 61 | | First boot through provisioning startup | 48.63 s | |
| | 62 | | Apply account and computer identity | 2.08 s | |
| | 63 | | Restart, log in, and start desktop capture | 43.92 s | |
| | 64 | |
| | 65 | The guest agent first answered at 33.10 seconds. Guest marker timestamps put |
| | 66 | identity changes at 1.83 seconds; the host's polling rounds this to 2.08 seconds. |
| | 67 | Two Kernel-General boot events confirmed the reboot. Explorer and guest capture |
| | 68 | were running, and the production browser displayed the personalized desktop. |
| | 69 | |
| | 70 | This is a disk template, not a saved-memory resume. `Rename-Computer` in the |
| | 71 | provisioning script is followed by a forced restart; the second boot enables |
| | 72 | capture and waits for the desktop. Avoiding that second boot is the next |
| | 73 | architectural opportunity. Cloning the disk itself is already inexpensive. |
| | 74 | |
| | 75 | By `ready`, QEMU recorded 474 flushes with 46.45 seconds of aggregate duration |
| | 76 | (98 ms per flush), 2.24 GB read, and 105 MB written. These operation durations |
| | 77 | can overlap and are not an additive wall-clock breakdown. The read cache does |
| | 78 | not accelerate synchronous HDD writes. This quieter-host result also has |
| | 79 | different cache warmth, so it does not isolate the benefit of fleet retirement. |
| | 80 | |
| | 81 | The restricted raw trace is |
| | 82 | `/var/lib/snowglobe/retirements/vms-20261020/windows-boot-trace.json`. |
| | 83 | The trace guest was shut down and removed afterward; the published preset and |
| | 84 | personal Mac remained intact. |