feat: support manual cleanup sandboxes by hittyt · Pull Request #446 · alibaba/OpenSandbox

hittyt · 2026-03-14T03:06:19Z

Summary

Add manual cleanup sandbox lifecycle support by treating omitted or null timeout as non-expiring sandbox creation.
Update Docker lifecycle handling, timeout limit validation, SDKs, tests, and docs to support nullable expiresAt and explicit cleanup semantics.
Document the controlled rollout path: upgrade SDKs first, then upgrade the server, then start creating manual-cleanup sandboxes.

Testing

Not run (explain why)
Unit tests
Integration tests
e2e / manual verification

Breaking Changes

None
Yes (describe impact and migration path)

Impact:

Lifecycle API responses may now return expiresAt = null for manual-cleanup sandboxes.
Older SDKs that assume expiresAt is always a timestamp may fail when they create, get, or list manual-cleanup sandboxes.

Migration path:

Upgrade SDKs/clients first.
Upgrade the server second.
Only then start creating sandboxes with timeout omitted or null.
Do not introduce manual-cleanup sandboxes into shared environments while older lifecycle API readers are still active.

Checklist

Linked Issue or clearly described motivation Proposal: Support non-expiring sandboxes / manual cleanup mode across Docker and Kubernetes #442
Added/updated docs (if needed)
Added/updated tests (if needed)
Security impact considered
Backward compatibility considered

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d45c49734f

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".

server/src/services/docker.py

hittyt · 2026-03-14T10:14:27Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7e35ecba27

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".

server/src/api/lifecycle.py

server/src/services/validators.py

hittyt · 2026-03-14T13:19:59Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cfeef4cca0

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".

sdks/sandbox/javascript/src/sandbox.ts

hittyt · 2026-03-15T01:08:47Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 639e9cb934

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".

server/src/api/lifecycle.py

sdks/code-interpreter/python/src/code_interpreter/adapters/code_adapter.py

sdks/sandbox/csharp/src/OpenSandbox/Options.cs

Pangjiping

LGTM. I'll be working with @Generalwin to get the kubernetes_service work done.

ctlaltlaltc · 2026-03-16T01:47:40Z

Thanks the great feature for long-running workloads.

Move _parse_and_accumulate_mount_refs() before the expires/manual-cleanup branch so that OSSFS ref counts are rebuilt for all sandbox containers on server restart, not just those with expiration timers. Made-with: Cursor

hittyt · 2026-03-16T08:08:27Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1aa1b164a5

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".

server/src/api/lifecycle.py

server/example.config.toml

TOML has no null literal; advise operators to comment out the key instead of setting it to null. Made-with: Cursor

ninan-nn · 2026-03-16T09:47:48Z

server/src/config.py

+        ge=60,
+        description=(
+            "Maximum allowed sandbox TTL in seconds for requests that specify timeout. "
+            "Omit from config to disable the server-side upper bound."


The description says "Omit from config to disable the server-side upper bound", but this field has default 86400, so omitting it still enables the cap. Could we align behavior and docs here (either default None, or update wording to reflect default cap)?

ninan-nn · 2026-03-16T09:47:50Z

server/example.config.toml

 log_level = "INFO"
 # api_key = "your-secret-api-key"  # Optional: Uncomment to enable API key authentication
 # eip = "1.2.3.4"  # Optional: External IP/hostname for endpoint URLs when returning sandbox endpoints
+# Maximum TTL for sandboxes that specify timeout. Comment out this line to disable the upper bound.


This comment says commenting out this line disables the upper bound, but with ServerConfig.max_sandbox_timeout_seconds defaulting to 86400, commenting it out does not disable the cap in practice. Please align guidance with runtime behavior.

ninan-nn · 2026-03-16T09:47:52Z

specs/sandbox-lifecycle.yml

            Sandbox timeout in seconds. The sandbox will automatically terminate after this duration.
-            SDK clients should provide a default value (e.g., 3600 seconds / 1 hour).
+            The maximum is controlled by the server configuration (`server.max_sandbox_timeout_seconds`).
+            Omit or set null to disable automatic expiration and require explicit cleanup.


timeout: null is documented here as manual-cleanup mode, but actual support is runtime/provider dependent (Kubernetes may reject when provider capability is false). Could we make this constraint explicit in the schema/docs to avoid client expectation mismatch?

Added necessary description.

ninan-nn · 2026-03-16T09:47:54Z

server/src/services/k8s/kubernetes_service.py

+        expires_at = None
+        if request.timeout is not None:
+            expires_at = calculate_expiration_or_raise(created_at, request.timeout)
+        elif not self.workload_provider.supports_manual_cleanup():


Implementation here correctly gates manual cleanup by provider capability. To keep API contract clear, please ensure lifecycle spec/docs explicitly state this is runtime/provider-dependent behavior (not universally available across all runtimes).

We are aiming to align this behavior for all runtime providers, however, k8s runtime implementation is not supported via this PR to avoid too large code change.

K8s runtime implementation should be provided via another PR.

Align default with documentation: omitting the key now truly disables the server-side upper bound instead of silently applying 86400. Made-with: Cursor

Kubernetes providers may reject null timeout when the workload provider does not support non-expiring sandboxes. Make this constraint explicit in the OpenAPI spec and schema description. Made-with: Cursor

Self-hosted runner's persisted build cache can serve stale class files from a previous branch, causing compilation errors when new methods (e.g. manualCleanup) are not found. Clean and disable build cache to ensure a fresh build every run. Made-with: Cursor

hittyt requested review from Generalwin, Pangjiping, jwx0925 and ninan-nn as code owners March 14, 2026 03:06

chatgpt-codex-connector bot reviewed Mar 14, 2026

View reviewed changes

server/src/services/docker.py Show resolved Hide resolved

chatgpt-codex-connector bot reviewed Mar 14, 2026

View reviewed changes

server/src/api/lifecycle.py Show resolved Hide resolved

server/src/services/validators.py Outdated Show resolved Hide resolved

chatgpt-codex-connector bot reviewed Mar 14, 2026

View reviewed changes

sdks/sandbox/javascript/src/sandbox.ts Outdated Show resolved Hide resolved

hittyt force-pushed the feat/non-expire-sandbox branch from cfeef4c to 639e9cb Compare March 15, 2026 00:57

chatgpt-codex-connector bot reviewed Mar 15, 2026

View reviewed changes

server/src/api/lifecycle.py Show resolved Hide resolved

Pangjiping added feature New feature or request component/server labels Mar 15, 2026

Pangjiping assigned jwx0925, Pangjiping and ninan-nn Mar 15, 2026

Pangjiping added the sdks label Mar 15, 2026

Pangjiping reviewed Mar 15, 2026

View reviewed changes

sdks/code-interpreter/python/src/code_interpreter/adapters/code_adapter.py Show resolved Hide resolved

sdks/sandbox/csharp/src/OpenSandbox/Options.cs Show resolved Hide resolved

hittyt requested a review from Pangjiping March 15, 2026 07:03

Pangjiping previously approved these changes Mar 15, 2026

View reviewed changes

hittyt mentioned this pull request Mar 16, 2026

feat: implement OSSFS storage backend in sandbox lifecycle #340

Merged

hittyt added 3 commits March 16, 2026 15:40

feat(lifecycle): support manual cleanup sandboxes

b8fe19e

fix(lifecycle): harden sdk compatibility and e2e stability

5e7627f

hittyt dismissed Pangjiping’s stale review via e934e56 March 16, 2026 07:57

hittyt force-pushed the feat/non-expire-sandbox branch from 639e9cb to e934e56 Compare March 16, 2026 07:57

fix(server): restore docker retry timedelta import

1aa1b16

test(java-e2e): relax mounted directory assertions

b6df0c6

chatgpt-codex-connector bot reviewed Mar 16, 2026

View reviewed changes

server/src/api/lifecycle.py Show resolved Hide resolved

server/example.config.toml Outdated Show resolved Hide resolved

hittyt added 2 commits March 16, 2026 16:42

fix(server): correct TOML null guidance for max_sandbox_timeout_seconds

a69cca3

TOML has no null literal; advise operators to comment out the key instead of setting it to null. Made-with: Cursor

fix(kotlin-sdk): add java manual cleanup builder

8da880d

ninan-nn reviewed Mar 16, 2026

View reviewed changes

hittyt added 3 commits March 16, 2026 18:04

fix(server): default max_sandbox_timeout_seconds to None

baed786

Align default with documentation: omitting the key now truly disables the server-side upper bound instead of silently applying 86400. Made-with: Cursor

docs(server): clarify manual cleanup is runtime-dependent

91802fc

Kubernetes providers may reject null timeout when the workload provider does not support non-expiring sandboxes. Make this constraint explicit in the OpenAPI spec and schema description. Made-with: Cursor

Conversation

hittyt commented Mar 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Testing

Breaking Changes

Checklist

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

hittyt commented Mar 14, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

hittyt commented Mar 14, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

hittyt commented Mar 15, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Pangjiping left a comment

Choose a reason for hiding this comment

Uh oh!

ctlaltlaltc commented Mar 16, 2026

Uh oh!

hittyt commented Mar 16, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

hittyt commented Mar 14, 2026 •

edited

Loading