I'd assume for "scales higher" on R2 you just means larger objects 😛
I'd assume for "scales higher" on R2 you just means larger objects 
Eyeah im struggling to word that rn
U
yeah it be like thatUMy personal path way of choosing is pretty much can I use a DO for this? Use a DO. Am I fine with higher cost / do I need/want faster responses and objects arent super large (over 25MiB)? Use KV. Else: Use R2.
UKV is very expensive comparative to R2.
UKV:
- $0.50/million get
- $5.00/million write/delete/list
- $0.50/GB-month
- $0.36/million get
- $4.50/million write/list
- $0.00/infinite deletes
- $0.015/GB-month
Uexpirations on KV do not count as a delete op
UAnd yeah the operation charges between the 2 arent that different (besides deletes). Its mostly the storage pricing. KV is normally faster but its also fairly easy to throw in R2 on cache and a cache hit would be pretty similar performance so I almost never see a reason to use KV 
U1 GB-s = 8 seconds of runtime for Workers.
A GB-sec basically means the number of seconds of runtime by the ram allotted. So since workers use 128MB of ram (or 1/8th a GB) 1 second would be 1/8th a GB-sec and 8 seconds = 1 GB-sec.
But there is also a really interesting thing to note there where you get charged the higher of the CPU usage multiplied by 8 or the wall time. Meaning if you use 1 second of CPU time but only 3 seconds of wall time you will get charged as if you did 8 seconds of wall time.
It's very hard to know how much CPU time you use but for the most part you will probably be charged by your wall time.
A GB-sec basically means the number of seconds of runtime by the ram allotted. So since workers use 128MB of ram (or 1/8th a GB) 1 second would be 1/8th a GB-sec and 8 seconds = 1 GB-sec.
But there is also a really interesting thing to note there where you get charged the higher of the CPU usage multiplied by 8 or the wall time. Meaning if you use 1 second of CPU time but only 3 seconds of wall time you will get charged as if you did 8 seconds of wall time.
It's very hard to know how much CPU time you use but for the most part you will probably be charged by your wall time.
UAlso for Workers pricing there is a new pricing model where you get charged only by CPU time and nothing else. (But DOs still get charged the GB-sec)
UYes you are always billed 128MB of ram even if you dont actually use it
UStarting October 31st, 2023 
UFor new users it will go away yes, im not entirely sure how that will play out for existing users if they force an immediate transition or what but at some point those will be gone for existing users also I assume.
For durable objects specifically though they still bill by what unbound is so I guess that kind of stays around
For durable objects specifically though they still bill by what unbound is so I guess that kind of stays around
UAh found what happens for existing users:
If you’re on our Workers Paid plan, you will have until March 1, 2024 to switch to the new pricing on your own, after which all of your projects will be automatically migrated to new pricing.
Uhaha all good 
UBoth of those syntaxes result in the exact same thing. But that specific error doesnt seem related to the incident occurring.
Are you sure you copied the bucket name correctly?
Are you sure you copied the bucket name correctly?
UAh interesting I didnt know you had to specify jurisdiction in the binding but I guess that makes sense
SHi all
I haver a question related with R2 pricing
Having a look into the metrics, looks like that whenever I access a file directly in the browser, it consumes some Class A & B operations
I thought this operations were for API operations access, but not to download a file
I haver a question related with R2 pricing
Having a look into the metrics, looks like that whenever I access a file directly in the browser, it consumes some Class A & B operations
I thought this operations were for API operations access, but not to download a file
KDownloading a file is class B, it’s still an API operation
KIf you use a custom domain and set up caching in front of it, you can conserve class B ops, serving from cache doesn’t hit R2
SI tried that, the first time the cache miss and the next ones hit the cache, but few minutes later if I try again, the cache miss again, reaching R2 and consuming class B ops
KYour file has to be accessed regularly to stay in the cache. Cloudflare’s CDN is very aggressive at evicting unused files
KYou can set up a cache rule to try to make it longer-lived, but they will evict unused files
KIf your file is popular or being accessed a lot it will stay in the cache and do the actual money saving there
SThanks Karew for your help. This makes the cost of my app difficult to predict ... will review other approaches
CClass B's are super cheap anyway (0.36/mil), If you were trying to predict your costs, I would just assume all will be uncached in your calculations and then go off real world data. It can vary, depends how spread out your users are as well as access frequency, etc.
but few minutes later if I try again, the cache miss again,This could also just be because you're hitting a cloudflare multi-colo pop, a cloudflare location made up of multiple datacenters, each with their own cache. Your file might not have been evicted, just your connection expired and you were routed to a diff colo
Mhi all, is there anyway to find out how much bandwidth you have used per month for your R2 bucket? im currently hosting a bucket that will have up to 200gb worth of files, and i would like to know if there is a way to find out how much you use.
KMthank you!
Mis there any other way? i keep getting this error when trying to execute the command to download denoflare
KI don't really know how to install it on Windows, sorry
Mah ok, thats all good
MHi all - we're currently testing out an R2 integration and are hitting intermittent CORS issues. One of our three engineers keeps getting CORS errors on a PUT presigned URL operation, myself and our other engineer have no issues at all. As far as I'm aware our CORS policy on the bucket is correctly configured.
MAny chance someone could help me debug?
KCompare the request headers of the failing requests with your CORS policy
MHi - yes we've done that. There's a fair few differences between the request and response headers across the two clients.
MSuccessful client -
MFailing client -
KWhat is the response body on the failing client?
MAfraid it's a cors error so the response isn't accessible
