After some of our client were complaining about severe packet loss, I started to dig into our configs we run on each of our sites.
We typically use auto-negotiation for everything as this ensures compatibility. However since we standardized on RouterOS 6.4 (main reason being able to use DNS name for OVPN instead of IP) we have picked up that Mikrotik 1100,1200, 1100AHx2 all seem to have an issue establishing a 1Gbps link to various Cisco devices (at our lab it’s a Cisco ME 3400) if you leave the Mikrotik on auto neg on ether1 to the Cisco which also has auto neg on it’s ports, you get a link but severe packet loss, if you force both sides to 100FD then everything works, but this doesn’t solve our problem as certain clients have over a 100M fiber link, if you force both sides to 1000FD then the Mikrotik doesn’t establish a link.
Upon further investigation, I started downgrading a test RB1100 with 5.20 and everything worked correctly, I incremented versions of powerpc up until 6.4, once I hit 6.4 I got this issue, and continued to get it up to 6.11.
Has anyone else experienced this negotiation issue between Mikrotik and Cisco? (I have seen plenty of threads here with no real outcome)
I noticed in the changelog as well, on 6.1
*) fix 1G linking with some Cisco devices (affects RB7xx, RB9xx, RB1100, RB2011, CCR);
However I only started to get this issue from 6.4+
Just tested 6.4 again with routerboot (also says 3.07 as latest)
not working
Also just tried 6.11 with routerboot 3.10
looks slightly different under ether properties, but still not working on forced 1000FD as was on 6.3 and older
Then I have no other idea than install 6.11, latest routerboot, prepare the situation, make supout.rif and send it to support. Anyway, sometimes could help netinstall and clearing the configuration.
We have had a very similar issue, but not quite the same.
Most Telco’s in our country provide circuits from Cisco PE devices and fix the speed/duplex at 100FDX, which unfortunately when we set this on the Mikrotik CCR will either not work at all, or even worse will work for a seemingly random amount of time before it stops working until the device is rebooted.
Our solution was to put a managed switch between the CCR and the telco, with the Telco port set to 100FDX and the CCR side set to Auto Negotiate. The CCR will happily autonegotiate with a HP Procurve, which can then negotiate with the Telco cisco or be set to 100FDX and works nicely.
I did raise a ticket with Mikrotik but had no spare hardware to help Mikrotik find the exact cause.
MattC86, please tell us exactly the full part number of the CISCO ME 3400 device. We will check if we could try to get it locally and check where could be the problem.
Next test will be to Auto both and see if that works, if that works, I’d say this is resolved, because even by IEEE standards, auto neg is the standard for 1Gbps neg.
