lets keep this conversation going. id like to share with hackernews, especially because of the other
lets keep this conversation going. id like to share with hackernews, especially because of the other recent events
HCan you send me the code for
kvworker, and make sure that the KV Namespace you screenshotted matches the ID you have in the wrangler.toml
HFeel free to do so, if you want
Dthe code in the worker..
Dcode in wrangler.toml
HAnd the URL for the namespace ends in
f51431a...?Df51431a4ff724d4892400056a9bb42f3
this is its the namespaceID
this is its the namespaceID
HSorry, not sure then...
Dyes, the namespace URL ends with
f51431a4ff724d4892400056a9bb42f3Dok, i figure it out
Dfirst we need to configure
[env.production] in wrangler.tomlwrangler deploy --env=productionD
ADoes anyone know how to resolve this error I'm trying to upload the mtls certificate
USounds like the cert you are passing isnt in a proper format what are you uploading in
AThanks, corrected and it worked
AHey all, I converted my worker to use smart placement over 2 days ago. I can see the placement status is set correctly
However, the dashboard says that there is still insufficient data for the request duration metric (there have been over 200 requests in the last 24 hours).
There is one thing I want to clarify, this worker is being invoked from a CF pages backend. So it's a CF worker calling this worker. Does that affect any metric collection? Should I not expect the request duration metric to show up eventually? Or is it just a waiting game?
"placement_status": "SUCCESS",However, the dashboard says that there is still insufficient data for the request duration metric (there have been over 200 requests in the last 24 hours).
There is one thing I want to clarify, this worker is being invoked from a CF pages backend. So it's a CF worker calling this worker. Does that affect any metric collection? Should I not expect the request duration metric to show up eventually? Or is it just a waiting game?
AIt's weird because I turned another worker into smart placement last night, it is still at the
"placement_status": "INSUFFICIENT_SUBREQUESTS", but it already has data in the graph.UHow many external requests do you make?
AThe worker calls D1 quite a few times per request I'd expect (it's a graphql API)
UI’d assume graphql probably combines them all into 1 query but afaik smart placement doesn’t do any of the “smart” logic when you use D1 and just places it near the D1 instance even if you don’t call it so not sure why it isn’t 
AHmm I see. But you're not sure if it would/wouldn't result in metric collection for the request duration graph?
AI guess, related question, does smart placement consider requests to R2 an "external request"? And place the worker appropriately?
JDoes anyone have additional information on how diagnostic_channel events are “forwarded” to tail workers? What is the mechanism of transit? I’m particularly interested in the security of this transmission.
AQuestion, do workers started by a cron trigger have the same subrequest limit (e.g they can only make 1000 subrequests per invocation?)
ECorrect
EInvocation is an Invocation regardless by which method its invoked, whether thats fetch or cron
ARight, and that subrequest limit is all inclusive with internal services too right? E.g if I make 1000 requests to KV, then I have no more bandwidth to make requests to something like R2
EI believe that's also correct
AOk, understood. Thanks 
Diirc you can 'get around' the subrequest limit by using service worker bindings
DI think we do that for one of our Workers that makes a crapton of requests at once
Aoh really? So those workers have no subrequest limit?
Dthey have the same 1,000 subrequest limit
Dif you make 1,000 subrequests to another service Worker that makes 1,000 subrequests, you now have a million requests in total triggered by 1 incoming HTTP request
Aoooh I see . Got it, very clever XD
EYeah so each invocation has the same limit but you can invoke a worker from another worker to get another 1k subreqs and "cheat" the system, though you do ultimately pay for more requests and if you recurse too deep you'll eventually run into an error to prevent infinite recursion
Dyou could even make a simple Worker that just takes in a list of URLs, and returns a list of responses from each lol
EAs an aside you can also make a Worker have a service binding of itself
Dif you go down through only 3 subrequests, aren't you already allowed to make 1 trillion requests? (1000 * 1000 * 1000 * 1000)
EIn theory yes but in practice you may find yourself falling victim to anti abuse systems if you try to launch 1 trillion subrequests
Dhah I would expect to be blocked at that scale anyway 
Dthe cost for 1 trillion subrequests in this example would be $300 right? (1000 * 1000 * 1000) * (0.30/1,000,000)
SShouldn't you also count in cpu time?
VWhy is this not working on /rewrite.html i get 404 error
KWhat's the route on the Worker?
Vdomainname/test.html
Vit is changing the url but i get 404 error on page
KThe Worker isn't involved with /rewrite.html if you don't have a route covering that, so the 404 isn't an issue from the Worker.
VWhat do you mean by that?
