The web UI no longer falls back to long polling (T-468) - #337
Merged
Conversation
…nnot demote a tab to a transport that dies on the next sleep (T-468) Root-caused from a laptop-sleep reproduction: the console line "WebSocket is closed before the connection is established" on the rejoin (_mounts=1) is phoenix.js's replaceTransport closing the still-connecting WebSocket when the 2.5-second longPollFallbackMs timer expires while Wi-Fi is coming back. The tab then runs on long polling until a full page load (the fallback is memorized in sessionStorage only when the WebSocket never passed a health check, which is why the key was null while the Network panel showed /live/longpoll). A long-poll session is a server process on the pod that first answered; it exits after 15 seconds without a poll, so every later sleep woke to 410 Gone, which phoenix.js raises as a transport error while the transport still counts as open, which LiveView shows as the server-error flash "Something went wrong!" — with nothing in the server log, since an idle shutdown is a normal exit. Three web pods without session affinity added a two-second PubSub hop to every poll. app.js drops longPollFallbackMs: a tab stays on the WebSocket and phoenix.js keeps retrying a slow open. The /live/longpoll route stays on the server for a client that asks for it. The deployment note says what an operator sees change: a network that blocks WebSockets now shows the reconnect flash instead of degrading quietly, so the load balancer must pass WebSocket upgrades. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NBmgYoYciJzbcxNYA9Z4d6
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.
Tracker: T-468.
Root cause, reproduced on a laptop sleep. The rejoin after wake logged
WebSocket is closed before the connection is established. That is phoenix.js'sreplaceTransportclosing the still-connecting WebSocket when thelongPollFallbackMs: 2500timer expires while Wi-Fi is still coming back. The tab then runs on long polling until a full page load (the Network panel showed/live/longpoll; the sessionStorage key was null because phoenix.js only memorizes the fallback for a tab whose WebSocket never passed a health check). A long-poll session is a server process on the pod that first answered and exits after 15 seconds without a poll, so every later sleep woke to410 Gone, which the client raises as a transport error while the transport still counts as open, and LiveView shows as "Something went wrong!". Nothing reaches the server log, since an idle shutdown is a normal exit. With three web pods and no session affinity, every poll also depended on a two-second PubSub hop to the owning pod.The change.
longPollFallbackMsis removed from the LiveSocket options inassets/js/app.js. A tab stays on the WebSocket and phoenix.js keeps retrying a slow open. The/live/longpollroute stays on the server for a client that asks for it explicitly.What operators see change. A proxy or network that blocks WebSockets now shows the reconnect flash instead of degrading quietly, so the load balancer in front of the web pods must pass WebSocket upgrades. A wake-up still shows the flash for the seconds the WebSocket takes to reconnect; that is a real reconnect, and its wording is the closed #336, to revisit separately.
mix precommitandmix cigreen; the bundle builds.🤖 Generated with Claude Code
https://claude.ai/code/session_01NBmgYoYciJzbcxNYA9Z4d6