I have 3 different CCR1072 all of them are getting RX overflow and RX control counters accumulating and all have .5% packet loss on the circuit. All the routers are moving under 300M on the ether1.

No other ports in any of my other Cloud Cores or Routers have the RX overflow accumulating.

Are the Ethernet ports not designed to handle traffic? Are they just for programming?

One thing I have also noticed is that Fast Path counters are counting on all ports but the Ether1 on the CCR1072

Response from Mikrotik

Hello,

Please check the block diagram:
https://i.mt.lv/routerboard/files/CCR10 ... 130622.png

ether1 is connected to a PCIe controller and it was designed that way since ether1 was only intended to be used as a management port.

rx-overflow counter is increasing when you have a faster device (or link) sending too much data into a slower interface. This is expected.

The solution is not to use ether1 for data links.

Best regards,
Arturs Z.

Not sure why it is not clearly defined as a management port. Maybe they could label it management instead of ether1 and take the 1G out of the CCR1072-1G-8S+ model number.

Wow. I see nowhere in the description or the block diagram that this port is intended to be used as "management-only." It looks like it should be a usable gigabit interface.

If this port isn't actually usable they shouldn't advertise it as such.

Does anyone know if this limitation of Ether1 also exists on CSR24 operating in Router Mode? I am seeing RX Overflow errors on Ether 1 with 200 mbps and 25k pps. Tried increasing Queue Buffers but problem persists.

If this is a different issue, what other testing/troubleshooting would you suggest?