Network / pulls slow in hosts inside the rancher

I’ve been looking around but I can’t find a good explanation or a way to look for at least a workaround, so I’ll try to post here and see if someone has experience for that.

I’ve been running a 1.6.15 rancher server happily for quite few months.
But a few days ago network operations and pulls from public/private hub started to show problems.

I’ve updated rancher to 1.6.17 and agents to 1.2.10. I’ve recreated connected hosts from scratch and updated their docker version from 1.1.13 to 17.06 and then to 18.02.
But no matter what I do now, every new host in my cattle has a slow connection.

To understand how much:

  • a normal host created in the same openstack cloud usually reaches Download: 916.59 Mbit/s
  • a new host in my rancher makes a speed test of Download: 4.15 Mbit/s

What should I do? Is there any component in the system stack I should patch? Why wasn’t it like that before?
Thank you all

It ended up with my company DNS working not well;
Using the ones from Google speeded up things again.