Page 1 of 1

RX Drops vs Pause Frames

Posted: Wed May 04, 2016 2:42 pm
by mhoppes
We have a WS-12-250B that is experiencing odd issues.

We have an EdgeRouter connected to a 1Gigabit fiber line which connects to the Netonix switch. The connection to the EdgeRouter and Netonix are through copper ports.

We have noticed the Netonix is increment the RX Drops counter even though the FC is turned on on this port. With FC turned on we sometimes see TX Pause Frames, but they aren't happening often, even though RX Drops is increment at a rate of about 3-4 a second.

Any thoughts on why this would be happening and why PFs aren't firing before the RX Drops happen?

Re: RX Drops vs Pause Frames

Posted: Wed May 04, 2016 2:48 pm
by sirhc
Well my first question would be is Flow Control indicated as active to the EdgeRouter on the Status Tab? We know UBNT has been working on Flow Control with their routers and some radios.

avatar.jpg
avatar.jpg (15.55 KiB) Viewed 7828 times


Port 1 shows Flow Control Active
Port 2 shows no Flow Control Active

Re: RX Drops vs Pause Frames

Posted: Wed May 04, 2016 2:50 pm
by sirhc
Rx Drops can happen for other reasons Matt not just because buffers over flowed.

I would also try and see if your modules work with Auto instead of 1G. Some modules may not allow Flow Control to work properly if not set to Auto. Remeber Flow Control is technically part of Auto Negotiations.

Re: RX Drops vs Pause Frames

Posted: Wed May 04, 2016 3:08 pm
by mhoppes
Flow Control does show as enabled on the Switch, and the Juniper (Which is the media converter between the edgerouter the fiber and the switch) also supports it.

We aren't using fiber modules, they are copper. There are two Junipers on each side of the fiber run (long haul fiber).

Re: RX Drops vs Pause Frames

Posted: Wed May 04, 2016 3:31 pm
by sirhc
Well I would investigate other causes of the drops other than suspecting they are being dropped because Flow Control is not preventing the buffer from overflowing.


RX Errors are typically caused by one or more of the following:
Positive values in RX-ERR counter mean that the NIC received malformed Ethernet frames from the transmitting switch port, and data integrity could not be validated during frame's cyclic redundancy check (CRC) . The root cause of this is usually a bad cable, or a bad interface on either the machine or the switch.

NIC speed / duplex mis-match with the connecting port on the switch/router.

High or critical performance rated IPS protections are set to Prevent or Detect.