Cloudflare

```fl=735f59
h=sheepit-main-weur-projects.sheepit-r2-binaries.win
ip=-
ts=1702503091.76
visit_scheme=https
uag=curl/7.81.0
colo=ORD
sliver=none
http=http/2
loc=US
tls=TLSv1.3
sni=plaintext
warp=off
gateway=off
rbi=off
kex=X25519```

```
fl=72f29
h=71c7a10d9126a4c692647ca88db1705f.r2.cloudflarestorage.com
ip=-
ts=1702503078.199
visit_scheme=https
uag=curl/7.81.0
colo=MSP
sliver=none
http=http/1.1
loc=US
tls=TLSv1.3
sni=plaintext
warp=off
gateway=off
rbi=off
kex=X25519
```

seems like Erisa was right on the money for this us based client

r2 gets way better performance probably because its pinned to a closer POP.

but that amount does not include stalled requests

they are just the ones that fail outright

so many question marks that need answering

image

22m r2 migration
14m rclone copy operation

appears like s3 traffic has half the latency that normal traffic has

or rather its simply on the enterprise priority

1.56% of requests just fail without ever reaching anywhere

if I had to guess only 95% of requests make it cleanly to r2

I added client side logging for network events

I have evaluated over 1.8 million log entries.
1.56% of requests to CF R2 fail outright
2.84% fail in some form during the transfer (connection reset, connection timeout etc).
so 4.4% of requests just die.

it's just okhttp3 on java doing the simplest get requests imaginable