https://aws.amazon.com/service-terms/ Section 82.3 specifically says you cannot cache most of the da
https://aws.amazon.com/service-terms/ Section 82.3 specifically says you cannot cache most of the data.
BNot yet. Internally we use v2 for all our tests so that's why we have v2 examples everywhere. We're happy to take contributions to add v3 examples.
FI could not find an example to store or retrieve data. Can you point me to an example?
FOr can I use my code from AWS and do the same, except for the region argument?
Fmy AWS v3 code.
KYou’ll need to specify the R2 endpoint as your bucketEndpoint in the S3 client constructor’s config
KThat’s all most the examples are doing - the code beyond that is the same as you’d use for S3, B2, GCS, etc for the most part
KI never made a V3 example since there was already the V2 one personally
Ktbh there was one SDK that rejected my R2 token since it’s a different length to the AWS tokens but I wouldn’t be able to tell you which it was since I never looked into how to get around it and just used another SDK
FOK I'll just try. BTW is it normal that with the V2 SDK, a listBuckets() takes about 2 seconds?
KK
KThat's an example
MThank you for this information
FThat looks like V3 to me. The import for V2 is
import S3 from 'aws-sdk/clients/s3.js';So your example is quite useful 
KYeah, it's a V3 example
KThere's already V2 examples in the docs
KIt was in reply to this
FAh thanks a bunch!
FI was able to store my first file.
BMind throwing that up in the docs so that we have both? Can be part of the JS page
KSounds good to me
KWill PR it in a bit
FFor me what I was missing compared to AWS are the
region: 'auto'https://${ACCOUNT_ID}.r2.cloudflarestorage.com\It works without the region to be fair - R2 accepts an empty string and
us-east-1autoautoThanks for the clarification. Still end
endpointYeah
KProbably just searching for it here and if it isn't mentioned, post it here haha
Kwould/should I use the S3 API for that?Yeah
KThere's a SDK for pretty much every language or tooling like rclone/aws CLI/etc so it should be pretty easy
Jrclone is the best!
BSame limits as S3
BNo. Amazon added those headers towards the end of our entry into open beta
BJust haven't gone & updated the docs
FThe following features are currently on the roadmap for R2 but not yet implemented.Pre-signed URLs are now implemented!
- Public Buckets
- Mapping a bucket to your custom domain
In the meantime, you can use a Worker to replicate the vast majority of functionality such as serving & caching public assets.
Take a look at the Demo Worker example here: https://developers.cloudflare.com/r2/examples/demo-worker/
Also, there's a great community project called Render that's made for R2:
https://github.com/kotx/render
Workers will not be required for making assets publicly available in the future, they are just an interim stop-gap.
J
B?
Jlooks like that landed in a dashboard release earlier today
BAh yeah
BI thought that had rolled out already
Llooks awesome, nice work
UIs there a similar limit with R2 where deleted objects can still "appear" in list operations like with KV? And can it take up to 60 seconds for an object to appear in the list?
BNo. R2 is strongly consistent. If an operation succeeds then the thing happened. It’s on our backlog to document the consistency guarantees specifically but GCS and S3 are very similar. Careful with 500 internal errors though. Just because an operation fails that way, it doesn’t mean nothing changed (eg maybe modification did happen or maybe it didn’t). I believe that’s only the case for 500 and random connection disconnects. All other errors typically have better defined semantics but I haven’t thought through if there’s some corner error case I didn’t think of where that isn’t the case.
B500 like that are also rare and more theoretical (like they could happen but in practice even 500 means it didn’t). Connection spontaneously dying without a response is more likely. But again also expected to be rare - if you’re seeing the connection just die without any response above 0.01% of requests (even on a 5 min spike), please let me know (obviously for s3 compat endpoint - workers bindings don’t have a visible concept of connection death).
BMany of our 500s are just us not translating more specific error codes correctly. I think I’m going to spend tomorrow trying to get that cleaned up.