I'm having an issue where I'm unable to copy files over 10k (approximately) in size. I'm running an application in a container on this machine and this application (CouchDB, specifically) can handle file transfers over HTTP. I'm not able to upload an attachment over 10k. But it's not related to the application or container as I try to scp a file to the machine. So it's not related to the application and it's not SCP specific, either. So I'm pretty convinced there is something going on at the network layer.

I've attempted transfers from different source machines as well, all with the same results. So the issue is almost definitely on the target side.

Destination machine is running CentOS 7. Not sure what else you would need to know, so please ask me for whatever relevant data would be useful.

Here is a partial output on an scp attempt. This is where it hangs for a long time until it fails.

This is a shot in the dark, but if you're uploading via HTTP and the target site uses php, you might want to take a look at the php.ini file on the target site. That file usually contains a limit on the size of files that can be uploaded.

Is it a managed switch? When things get funky around 10K, I suspect jumbo frames which occur around 9018 bytes. Jumbo frames is a way to increase the MTU from 1518 to 9018 to increase speed. If the switch is unmanaged, then that isn't it.

I don't think that's it. If it was only the application, you'd probably have something here. But scp is also giving me grief at the same file size. My instincts tell me that it's somewhere in the TCP Stack.

The MTU on source and destination are all set at 1500. Also, the target is a VM on an Antsle. Other VMs on the Antsle do not have this problem. So the switch is unlikely the issue.

Try running a tcpdump while you do the upload.
If saved as a pcap, makes it easy to open in wshark