But if we use storage operations on the waitUntil, we will be able to make a response without actual

But if we use storage operations on the waitUntil, we will be able to make a response without actually wait.

WWallacy But if we use storage operations on the waitUntil, we will be able to make a res...

kenton•8/3/21, 11:06 PM

Ah. Yes, deferring code to run after the response has already been returned is inherently unreliable, because in the case of a failure, no one finds out. E.g. if the power goes out, the task might never execute, but since the user already received a response, they won't know. Output gates can't help with this, since they can't predict what you're going to do in the future after you return the response. Generally, I don't recommend doing this, except maybe for stuff that is fine to be unreliable, like logging.

Kkenton lol just came here to post that, you beat me

codefrau•8/3/21, 11:08 PM

Very good writeup about an exciting design. Can't wait to test. Just bogged down with other stuff

asp•8/4/21, 9:15 AM

Does anyone have numbers on roundtrip time to Durable Objects living on the other side of the world?

Kuririn•8/4/21, 1:50 PM

what is the difference between idFromName() and idFromString() and how long value (max. length) I could put there?

KKuririn what is the difference between idFromName() and idFromString() and how long valu...

kenton•8/4/21, 1:55 PM

specifically parses the output of . It won't work with any string that wasn't specifically created by stringifying an ID from the same namespace.

idFromName()

idFromName()

takes any string you want and returns a unique ID corresponding to it.

kenton•8/4/21, 1:56 PM

in retrospect, the naming could have been better.

Kkenton `idFromString()` specifically parses the output of `id.toString()`. It won't wor...

Kuririn•8/4/21, 2:05 PM

thanks!

Kkenton in retrospect, the naming could have been better.

asp•8/4/21, 2:27 PM

Aasp Does anyone have numbers on roundtrip time to Durable Objects living on the othe...

codefrau•8/4/21, 3:48 PM

LAX to KIX is ~ 130 ms, LAX to FRA ~170 Where are you? I can send you a link to try

Ccodefrau LAX to KIX is ~ 130 ms, LAX to FRA ~170 Where are you? I can send you a link t...

asp•8/4/21, 3:52 PM

I'm in Denmark, so I hit CPH by default. Though, due to VPN on my PC, I mostly connect to AMS.

asp•8/4/21, 3:54 PM

My latency to CPH is ~10 ms and AMS is ~25 ms.

codefrau•8/4/21, 3:59 PM

Try these for latency testing:

US

US

https://croquet.io/cloudflare/?EDHEob1TnShyWTdrRoe-GCIdl4E9x7Zz6qqJg7WFnpS

JP

JP

https://croquet.io/cloudflare/?3honQOd0yTj4F0-2-oaIFpP4D4n4S-9S3joiVfLY4nm

EU

EU

https://croquet.io/cloudflare/?ihlwFIj1Z9-BtUfoNAEzD0ItTEZl1qsZrpR8n_BNx2V
Click "test latency" a few times, and maybe reload to make a new request.

Aasp My latency to CPH is ~10 ms and AMS is ~25 ms.

codefrau•8/4/21, 4:04 PM

Looks like your VPN is slowing down things maybe? FRA should be a lot less.

asp•8/4/21, 4:05 PM

Oh, I'm on a mobile connection right now - that's definitely why. Will test it in ~30 min when I'm home

codefrau•8/4/21, 4:05 PM

Oh that is a problem too of course. In LA I sometimes get single digit latencies

(on fibre at home)

asp•8/4/21, 4:06 PM

on fibre at home

I "only" have 1000/500 over COAX (DOCSIS 3.1) :/

asp•8/4/21, 4:07 PM

Speeds are great, but latency doesn't get lower than ~10 ms to anywhere

asp•8/4/21, 4:15 PM

@codefrau I launched the links on my PC now (with VPN)

Aasp @codefrau I launched the links on my PC now (with VPN)

codefrau•8/4/21, 4:16 PM

So that's the connection from Amsterdam then - can you try without VPN? 50 ms to FRA still seems like a lot

Ccodefrau So that's the connection from Amsterdam then - can you try without VPN? 50 ms to...

asp•8/4/21, 4:18 PM

Done

codefrau•8/4/21, 4:18 PM

32ms, that looks better

asp•8/4/21, 4:19 PM

So 32 ms is roundtime from my PC to FRA via CPH?

codefrau•8/4/21, 4:20 PM

yes

codefrau•8/4/21, 4:20 PM

the worker is in CPH and the DO in FRA

asp•8/4/21, 4:22 PM

Doesn't 50 ms seem seem alright then, if it's from my PC -> Amsterdam (VPN) -> CPH (Worker) -> FRA (DO)?

codefrau•8/4/21, 4:25 PM

I guess it is – I am spoiled by having the worker right here next to me

asp•8/4/21, 4:26 PM

I don't think there's anything to do about the ~10 ms latency added just to get out of the house

asp•8/4/21, 4:26 PM

this is COAX after all

asp•8/4/21, 4:28 PM

Anyways, thanks a lot for the help

codefrau•8/4/21, 4:28 PM

You're welcome

Aasp Doesn't 50 ms seem seem alright then, if it's from my PC -> Amsterdam (VPN) -> C...

codefrau•8/4/21, 4:49 PM

Just to correct this - your connection from Amsterdam did not go via CPH. There is a colo in AMS, so that’s where the worker was.

codefrau•8/4/21, 4:49 PM

(the “via” item in the latency display is the worker colo)

asp•8/4/21, 4:53 PM

Ahh, alright - that's quite slow then, as I get 26 ms latency when running a Speedtest on the VPN. Cloudflare does peer with M247 in Amsterdam, right?

asp•8/4/21, 4:56 PM

Hmm, pinging

1.1.1.1

1.1.1.1

(on VPN to M247 in Amsterdam) gives me a latency of 28 ms, so perhaps the route between AMS and FRA is just slow for some reason?

ItsWendell•8/4/21, 6:01 PM

Is it possible to move an web-socket connection from the worker to a durable object?

IItsWendell Is it possible to move an web-socket connection from the worker to a durable obj...

codefrau•8/4/21, 7:04 PM

yes, you pass the websocket request (with the "Upgrade: websocket" header) to the DO via

stub.fetch(request)

stub.fetch(request)

Kuririn•8/4/21, 9:15 PM

When I have durable object, let's call it xyz I must use fetch to communicate with it?
or could I call it's functions directly?
example: xyz.storage.get('key');

KKuririn When I have durable object, let's call it xyz I must use fetch to communicate wi...

codefrau•8/4/21, 9:20 PM

Yes, fetch only

Ccodefrau Yes, fetch only

Kuririn•8/4/21, 9:20 PM

ok, thanks

Vvans163 seeing these today "Error: Internal error in Durable Object storage write caused...

kenton•8/5/21, 2:09 AM

It looks like we have some sort of bug where some errors like this happen while we roll out an update to the workers runtime. We are looking into it. (FWIW, whenever you see "internal error" that means we think it's our bug, not yours, and the error has already been reported to our internal monitoring.)