Skip to content

Mvm docs - #704

Merged
breardon2011 merged 11 commits into
mainfrom
mvm-docs
Sep 3, 2026
Merged

Mvm docs#704
breardon2011 merged 11 commits into
mainfrom
mvm-docs

Conversation

@breardon2011

@breardon2011 breardon2011 commented Sep 2, 2026

Copy link
Copy Markdown
Contributor

Adds a Sandboxes · MicroVM tab (21 pages) beside the existing Sandboxes docs, which are untouched and still first — customers can read both while preparing to move.

Written against behaviour verified on dev. The migration-relevant findings:

  • 8h hard lifetime, counting hibernated time. endAt is measured from host launch, so read it rather than computing it
  • No 16 GB tier — 1/4/8 GB carry over, and memory can't change after launch
  • Disk fixed at ~16 GB, down from 20 GB; diskMB is accepted but ignored
  • Templates carrying a rootfs are refused — most existing ones do, so they need rebuilding
  • Checkpoints are filesystem-only; restore is a fresh boot, not a resumed process
  • Secrets proxy moved into the guest as root — same sealed-token model, larger blast radius if code escalates inside a sandbox

Also adds deploy/microvm/SIZE-TIERS.md, the prerequisites for publishing a size tier.

Microvm branch WIP here, mostly complete: #703

…re to migrate — 21 pages covering lifecycle, the 8-hour ceiling and endAt, filesystem-only checkpoints, sizes, the in-guest secrets tradeoff and the supported/unsupported split, with the existing Sandboxes docs left untouched and still first
…nly the default pooled) instead of implying any size can be published, and rewrite the secrets page as an explicit before/after — the substituting proxy moved from the VM host into the guest as root, which changes only the blast radius of an in-sandbox privilege escalation
…easurable from inside the guest (nproc reports the host, not the share) so the numbers were unsubstantiated; say instead that CPU follows memory and that cpuCount and diskMB are accepted but are not controls on this runtime
…it cannot reach the platform's launch call, and it opts the create out of the warm pool into a cold launch for no benefit
…rying a rootfs image are refused outright (which is what most existing templates are, so they need rebuilding), and disk is fixed at ~16GB with diskMB accepted-but-ignored, down from a 20GB default; both added to the checklist, FAQ and comparison tables
…ublish.sh and microvm_common.go, and correct the customer-facing range to the 2-8GB we can actually offer — publish.sh treats 2048 as the vCPU floor and warns below it, so listing 512/1024 as available sizes was wrong
…runtime's tiers — 1/4/8GB carry over, 2GB is new, 16GB has no equivalent; the sub-2048 publish warning is expected for the 1GB tier since the current 1GB tier is already best-effort CPU, so it is not a regression
@mintlify

mintlify Bot commented Sep 2, 2026

Copy link
Copy Markdown

Preview deployment for your docs. Learn more about Mintlify Previews.

Project Status Preview Updated (UTC)
opencomputer 🟢 Ready View Preview Sep 2, 2026, 1:38 AM

💡 Tip: Enable Workflows to automatically generate PRs for you.

…roVM tab, per review — v2 mirrors v1's page structure one-for-one under v2/**, drops all infrastructure vocabulary in favour of 'v2 with breaking changes', keeps removed features as honest not-available pages so the structures still match, and adds a dedicated migrating-from-v1 guide
… entry in navigation.versions, so v2 now leads and v1 stays reachable from the switcher
@breardon2011
breardon2011 merged commit dae4123 into main Sep 3, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant