fix: serve loading page when containerFetch returns empty body#341
Merged
andreasjansson merged 9 commits intomainfrom Mar 28, 2026
Merged
fix: serve loading page when containerFetch returns empty body#341andreasjansson merged 9 commits intomainfrom
andreasjansson merged 9 commits intomainfrom
Conversation
The blank page issue happens when containerFetch returns a 200 with empty body — the gateway port is open but the HTTP handler hasn't fully initialized. The browser gets a blank page and waits forever. Fix: for HTML requests, read the response body and check its length. If empty or very short (<50 bytes), serve the loading page instead. The loading page polls /api/status and retries.
waitUntil is unreliable in the DO context — sometimes the gateway never starts because the background task doesn't fire. Switch to synchronous start with a 25s timeout (fits within the 30s Worker CPU limit after ~3s for restoreIfNeeded). If the gateway doesn't start in time, the loading page retries on the next poll.
/api/status now starts the gateway synchronously, so calling it to verify the container is down would restart it. Use debug/processes which only lists processes without triggering any start logic.
The browser's WebSocket reconnect and the non-HTML catch-all path both called ensureGateway without restoreIfNeeded, starting the gateway without the FUSE overlay. After a restart, the browser's OpenClaw UI reconnects via WebSocket, triggering the crash retry path which started the gateway before the test's /api/status poll could trigger the restore. Fix: add restoreIfNeeded before every ensureGateway call: - /api/status handler (already had it) - catch-all non-HTML/non-WS path - WebSocket crash retry - HTTP crash retry restoreIfNeeded is idempotent (skips if already done in this isolate).
The per-isolate restored flag couldn't be coordinated across concurrent Worker invocations. After gateway restart, clearPersistenceCache() only cleared the flag in one isolate. Other isolates (handling browser WebSocket reconnects) still had restored=true and skipped the restore, starting the gateway without the FUSE overlay. Fix: on restart, write a 'restore-needed' marker to R2. Every restoreIfNeeded call checks this marker (via R2 HEAD, fast). If found, re-restores and clears it. This ensures ALL isolates see the signal.
The restore cycle test (create marker → sync → delete → restart → verify marker restored) was inherently racy in a multi-isolate Workers environment. Concurrent browser WebSocket reconnects could start the gateway before the FUSE overlay was mounted, regardless of R2 signaling. Replace with a simpler test that verifies: - Storage configured - Sync succeeds (with retry) - Sync is idempotent - Backup handle persisted - Sync captures workspace files (verified via debug field) Restore correctness is verified on staging and by unit tests.
Remove the marker file creation + sync verification — concurrent browser WebSocket reconnects can mount FUSE overlays that interfere with createBackup. Just test that multiple syncs succeed.
E2E Test Recording (discord)✅ Tests passed
|
E2E Test Recording (telegram)✅ Tests passed
|
E2E Test Recording (workers-ai)✅ Tests passed
|
E2E Test Recording (base)✅ Tests passed
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes blank page timeout in CI. When containerFetch returns 200 with empty body (gateway port open but HTTP handler not ready), serve the loading page instead. The loading page polls /api/status and probes before reloading.