This usually means that there are network problems somewhere between your agent and the service where you get interrupts.
Starting with agent 2.x, we are actively trying to work with network conditions. We break each file into pieces (4 MB), open an HTTP stream and pass this fragment of the file to the service. Each fragment of the file is set for a long time (10 minutes)
We have attempted to retry for retrying.
Backoff 5.16 seconds before attempt '2' chunk '1'
Then, if we still cannot load the piece, even with the help of attempts, we put this file in the repeat queue for another pass.
So, as you can see, we are trying very hard.
The more network interrupts and the larger the file, the greater the chances. In addition, if the download speed slows down a lot, it can lead to crashes during block loading (4 MB in 10 minutes - a long time), but in your case it looks like the stream / network is interrupted.
You must check the network. Something is interrupted between your machine and the service.
Another experiment that needs to be tried is a private agent on a private virtual virtual machine (e.g. azure). This gives you a faster cloud network at boot time, as well as private agent management (install the right software, incremental source and packaging, etc.).
Do you see the repetitions that I am describing? I also forward this to our developer agent. I will edit if anything else is added.
source share