yup i had done something dumb in my code while trying random stuff to get the object to work. its all fixed now. thank you so much for the help

Btw just curious do websockets get charged per message sent or only based on how long its connected? Whenever I look at the metrics any websocket messages dont show up as requests so I assume its charged based on connection time?

I am still getting some

Network connection lost.

errors (sometimes ~400 errors a day) for requests interacting within DurableObjects. The stub is constructed every time, and its happening pretty random. Is there something known how to fix it or shall I just retry the DO request when it happens? Unfortunately I didn't have any luck replicating it.

You should definitely add retry logic, since DO requests may cross the globe, some amount of connection loss is to be expected. That said, we're always working on improving that. Do you know what PoP you're hitting to access your normal worker?

colo

from here: https://www.cloudflare.com/cdn-cgi/trace And when you say ~400 in a day, out of how many requests total is that?

Sometimes it seems the DO is stuck, as repetitive requests fails as well. I will try to add some retry logic today, it should definitely narrow down potential bigger issue. 🙂 That’s a good point, it’s from ~500k daily requests

@User Do you recreate the client stub (calling

namespace.get(id)

) for every request, or do you have one stub you're reusing across many requests?

recreating the stub for each request, using

idFromName

ok... that's the right thing to do.

trying to improve the error logging a bit, so hopefully gonna give you more context soon

The documentations on durable objects + websockets is still pretty spotty, I see this under DOs > Full documentation for WebSockets will be coming soon, but for now check out this heavily commented example chat application that runs in Durable Objects to see how it works. Unfortunately, I'm trying to do something that isn't reflected in the chat app - namely, have the DO act as a client to a websocket server abstracting over native code running on another machine, and route requests from workers to the backend via the DO websocket client connection.

I could refactor my system to make the DO a WS server, perhaps, but everything's shaping up with the backend acting as the websocket server, exposed via a cloudflared tunnel.

Is it possible for a DO to initiate a websocket client connection and proxy requests from workers to a third party websocket server? Seems like it should be, but I am at a bit of a loss with the DO + WS client mode combo.

Something I find very needed working with DOs/workers is support for https://developer.mozilla.org/en-US/docs/Web/API/GlobalEventHandlers/onerror. A GlobalEventHandler for exceptions. So any crash anywhere, at any level, can be shipped off to our logging platform. (instead of having maticulous webs of try/catch/awaits, sending errors 1 way for websockets, sending errors another way for http requests.

How do you do a "Expect: 100-continue" inside a worker using fetch? I have an endpoint that wants to use this pattern

Basically the fetch gets the response headers, if the status code == 100, the fetch needs to start uploading the body. I dont see support inside fetch for this

ok wow i didnt know about that until now, thank you, i hate error handling

DMs open, very interested to hear if this is tractable.

i think it should be possible, if you get it done and wanna share can you DM lol im curious too

onerror

does look nice, but I'm not sure why you need meticulous try/catches instead of it, why not a top level try/catch of your own? https://gist.github.com/bretthoerner/79cd3f823c9a7e2fa8adfb5a82d857d2

this does not work for backaround async awaits. (example calling a async function without awaiting for background task/sync)

so you get this messy stub for every function

Yeah I guess you’d have to wrap each promise in a wrapper function, but that same wrapper could be used for all of those, at least

Anybody know of any current issues/outages with DO? I'm just getting "internal error" when "fetching" on the DO stub.

same here. getting "internal error" when trying to fetch

😢

Real shame - because it's the last day of a hackathon I'm competing in, and we needed to record videos of our "creation"... and nothing's working :/

Sadly we’ve had to throw in the towel with the hackathon. We unfortunately left it to the last day to record the video - our bad. I guess that’s how it goes with beta software 😅

What's your nearest CF colo? https://cloudflare.com/cdn-cgi/trace will tell you

I also got 10.1k events in Sentry from around that time with "Internal Error", cant tell which colos were affected exactly

I can give you affected DO ids though, if there are no migrations in place yet it should tell you which colo was prob affected