fix: restart the dev proxy server whenever it closes
#317
Conversation
🦋 Changeset detectedLatest commit: 0343636 The changes in this PR will be included in the next version bump. This PR includes changesets to release 1 package
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
threepointone
force-pushed
the
restart-expired-sessions
branch
from
95da320
to
5ec1192
Compare
There was a problem hiding this comment.
Not having any way of automatically testing this, I assume that you have done exhaustive manual testing ;-) I can't think of any reason why simply recreating the remote client would not work.
What happens if you are in the middle of debugging? What is the developer experience? Does the Worker appear to crash and restart?
I left a couple of NIT comments, which you can consider.
| @@ -119,6 +135,11 @@ export function usePreviewServer({ | |||
| const remote = connect(`https://${previewToken.host}`); | |||
| cleanupListeners.push(() => remote.destroy()); | |||
|
|
|||
| // As mentioned above, the session may die at any point, | |||
| // so we need to restart the effect. | |||
| remote.on("close", retryServerSetup); | |||
There was a problem hiding this comment.
Very clean solution.
When we run `wrangler dev`, the session that we setup with the preview endpoint doesn't last forever, it dies after ignoring it for 5-15 minutes or so. The fix for this is to simply reconnect the server. So we use a state hook as a sigil, and add it to the dependency array of the effect that sets up the server, and simply change it every time the server closes. Fixes #197 (In wrangler1, we used to restart the whole process, including uploading the worker again, making a new preview token, and so on. It looks like that they may not have been necesssary.)
threepointone
force-pushed
the
restart-expired-sessions
branch
from
5ec1192
to
0343636
Compare
|
We don't do breakpoint debugging at the moment, so there's no crash visible. I assume we'll have to work on this after we land breakpoints in the debugger. One thing that I think will break, is websocket connections. But that's possible even without dev, and folks should write code that's resilient to websockets dropping and reconnecting. We'll probably have to iterate on this more in the future when we're working on developing multiple workers, but this is a good place for now, at least for parity with wrangler 1. |
When we run
wrangler dev, the session that we setup with the preview endpoint doesn't last forever, it dies after ignoring it for 5-15 minutes or so. The fix for this is to simply reconnect the server. So we use a state hook as a sigil, and add it to the dependency array of the effect that sets up the server, and simply change it every time the server closes.Fixes #197
(In wrangler1, we used to restart the whole process, including uploading the worker again, making a new preview token, and so on. It looks like that they may not have been necessary.)