Yup. Should be up now

Working for me as well!

keeping notes/bugs about r2 here: https://gist.github.com/johnspurlock-skymethod/6027c81953f81aa535d889d86a1061ba

There may be an ongoing outage for some pre-existing buckets until tomorrow or Friday.

The PUT size limit also seems to apply to Worker bindings. I'm getting

Error: internal error

when trying to PUT a 1.3 GB file.

That limit be in front of your worker

Unless you're generating a 1.3 GB stream?

Albert I'm also seeing your bucket is one of those unfortunate pre-existing buckets

Try a different bucket for now

(Trying to remediate more buckets)

Looks like mine is on that list too (seeing 500s on

PUT

) 🙂 No big deal - not using it for any prod workloads, but excited to see those perf improvements 😄

Try now?

@albert feel free to try too

In a worker, if you're doing a subrequest to fetch the 1.3GB file and feeding that into R2, that should work. Let me know if it doesn't

Hmm PUT was successful now, but all GETs are failing for files. My keys are stuff like

2022/03/abc.png

DM me your account tag & bucket name

I didn't see any GETs for

2022/03/abc.png

in the logs

Seems it's working now and I was maybe just impatient - weird. Thanks for checking though 🙂

Perf competition for internet points: Who can send the most number of successful concurrent reads/s into a single bucket? Who can send the most number of successful concurrent reads/s into a single key within a bucket? Who can send the most number of successful concurrent writes/s into a single bucket? I have no way of actually verifying any of the numbers posted so honor system.

Challenge accepted 👀

Challenge impossible 😅

Time to fire up a few hundred "All Regions" 10 second health checks 🙃

As a point of reference, the current winner is my single-threaded node instance from my questionable Comcast connection seemed to be able to do ~1500 reads/s.

Writes were more moderate at ~200-300 writes/s for the bucket which indicated there might be a bottleneck on our end

I think we should decide on a specific object size.

What do you propose?

0 byte will be the fastest

But if you want to set up a 10 MB and 100 MB tier go for it

Keep in mind you probably want to avoid multipart for now

It's on my backlog to revamp that part of the code to work better