Skip to content

Sync latest upstream and E2B optimization with tracing into beta - #24

Merged
sam2tom merged 35 commits into
betafrom
codex/e2b-beta-main-sync-20261011
Oct 11, 2026
Merged

sam2tom merged 35 commits into
betafrom
codex/e2b-beta-main-sync-20261011

Conversation

@sam2tom

@sam2tom sam2tom commented Oct 11, 2026 •

Copy link
Copy Markdown

Summary

Synchronize beta with the latest MiniMax main and the updated E2B optimization and startup tracing branches. The integration preserves the fork's production deployment workflows, lease diagnostics, journal failure handling, and Environment file rejection logs.

  • Keep native E2B pause/resume and fast startup on the current shared Sandbox Provider protocol, including checkpoint compatibility, durable operation fencing, and retained-state capacity accounting.
  • Preserve the startup-to-first-output trace correlation across Core, Runtime, and E2B helpers.
  • Upgrade the integration to schema 99 and Runtime wire 0.16.0; extend the beta upgrade regression to verify retained state preservation, checkpoint constraints, and refusal to downgrade across multiple historical allocations/devices.

Upstream contributions remain in MiniMax-AI#642 and MiniMax-AI#643. Fork-only production changes are not sent upstream.

Validation

  • Complete integration suite: 760 passed / 12 skipped; the operator executable skip was subsequently verified separately. Native daemon/model qualification remains outside these local fixtures.
  • Execution package: 322 passed. Deployment persistence: 105 passed, with two host-history failures also reproduced on the exact upstream baseline.
  • Final retained-state increment: 58 race tests passed, including six cases using Core's real serialized compute state; generated SQL verification passed.
  • Isolated PostgreSQL 91-to-99 upgrade fixture preserves all 107 allocation records and 17 suspended records; verifies schema 98 uniqueness and downgrade refusal, and schema 99 checkpoint object constraints.
  • Whole-repository compilation and bilingual documentation checks passed.
  • Forced full workflow_dispatch CI 38105685214 passed on exact commit 7edd85ae90636c42d5bf2cfeba71cfade4e675d2, including Linux helper, all backend/integration shards, four Web E2E shards, Harness, examples, lint, compose, and Linux/macOS/Windows native checks. Local macOS limitations and baseline flakes remain recorded separately.

Production rollout

Merging beta does not deploy production. The schema 91-to-99 production upgrade requires a verified database/state recovery point and a safe plan for the existing old-wire suspended guests. This PR does not authorize deleting or waking those guests and does not include a production deployment.

sam and others added 30 commits October 4, 2026 13:02
* Preserve native sessions across compute replacement

* Use allocation storage contract during checkpoint restore

* Normalize cancellation while waiting for compute

* Attach administrator audit source to deployment updates
* Queue sessions for available compute

* Align hosted creation test with deferred placement
* Preserve Runtime PATH in login shells

* Keep PATH literals aligned in translated image guide
* Classify workspace storage exhaustion as unavailable

* Run storage exhaustion assertions in native platform CI
* Restore retained checkpoints across compatible nodes

* Align checkpoint fixtures with explicit restore admission

* Wake checkpoint destinations and settle canceled restore preparation

* Close expired restore requests before native dispatch

* Accept checkpoint qualification in node health JSON

* Document checkpoint compatibility during node maintenance
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 11, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-11T02:40:08.173545Z 7edd85a PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7edd85ae90

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment on lines +177 to +179
next, err := w.ProvisionEnvironment(ctx, environment.TenantID, environment.ID, owner.ProviderKey)
if err != nil && next.ID == "" && !errors.Is(err, placement.ErrNodeUnavailable) && !errors.Is(err, placement.ErrNodesPreparing) && !errors.Is(err, deployment.ErrAllocationConflict) {
return err

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Treat an unassigned queued placement as transient

In node-backed mode, ProvisionEnvironment calls syncNodes, which processes only the scheduler's current bounded demand page, and then returns ErrExecutionUnavailable when this Environment still has no placement. When the replacement demand is on a later page—or outside the current fixed sweep horizon—next.ID is empty and this condition returns the error immediately, causing a live file request to fail with 503 instead of continuing the existing wait loop until the common scheduler reaches it. Handle this no-placement outcome like ErrNodeUnavailable/ErrNodesPreparing so retained Environments do not bypass or fail merely because of their queue position.

AGENTS.md reference: AGENTS.md:L44-L44

Useful? React with 👍 / 👎.

@sam2tom
sam2tom merged commit 073462f into beta Oct 11, 2026
50 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.

3 participants