[hectorvent]

Summary

When restart had removed the old container and the new one failed, the settings of the instance were gone with it and nothing said how to get them back.

• If the new container did not come up, stderr now prints a start command with the same container, port, image, persist dir, services and namespace.
• If the new container is running but did not become ready, stderr says to check logs and run wait, because start refuses a running container.
• Both name the tree (floci aws), a typed --endpoint is kept, and --persist prints the directory a local daemon would mount.
• Applies to every tree through the shared RestartCommand. Exit codes and -o json stdout are unchanged.

Type of change

• Bug fix (fix:)
• New feature (feat:)
• Breaking change (feat!: or fix!:)
• Docs / chore

Product trees affected

• AWS (root tree)
• GCP (commands/gcp/)
• Azure (commands/az/)
• N/A (shared infrastructure, docs, CI)

Checklist

• mvn test passes locally
• New or updated tests added (or explained below why not needed)
• README.md updated (if a command or flag changed)
• Native binary verified (mvn package -Pnative), required if Jackson serialization or dependencies changed
• Commit messages / PR title follow Conventional Commits: a feat/fix/perf subject becomes the release note, so CHANGELOG.md needs no hand edit
• I have no more than 2 open, non-draft pull requests and no more than 4 open pull requests in total (drafts included) in this repository (maintainers and dependency bots are exempt).

Tests

RestartRecoveryTest covers a failed docker run in text and JSON, a failed stop, and a readiness timeout. 760 tests, 0 failures. Ran the fat JAR against real docker with nginx:alpine as a container that runs but never answers. Native build and smoke pass on linux/amd64 and darwin/arm64 with this change merged alongside the other open work; the smoke script does not exercise restart. OCI is covered too: the template has no box for it.

[greptile-apps]

Confidence Score: 4/5

[Medium impact] The change affects restart recovery across the four product trees.

The remote-host recovery issue still needs a complete fix before merging.

Findings

1.  Recovery waits on wrong host ▶

Summary

Adds recovery advice when restart fails after removing the old container.

• Failed restarts now show the right way to recover.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Resolve and check settings] --> B[Stop and remove old container]
    B --> C[Save recovery command]
    C --> D[Start replacement]
    D --> E{Startup succeeded?}
    E -->|Yes| F[Return success]
    E -->|No| G{Replacement running?}
    G -->|Yes| H[Suggest logs and wait]
    G -->|No| I[Print start command]

Loading

_{Reviews (3) · Last reviewed commit: "chore(skill): bump the skill version to ..." · Reviewed by Greptile}

[greptile-apps]

Recovery waits on wrong host

The printed wait command drops the endpoint used by restart. If a remote Docker instance was restarted with --endpoint http://build-host:4588 on host port 14588, the embedded wait checks build-host:14588, but the suggested command checks localhost:14588. It can keep timing out after the remote server becomes ready. Include the endpoint that start actually waited on.

Knowledge Base Used: Environment lifecycle operations