Hello,

I have a CRS310-1g-5s-4s+ running 7.6 (I've also tried v7.7rc4) that is suffering from a major port flapping issue when using either UF-MM-10G (Ubiquiti) 10G SFP+ modules or 10Gteck SFP+ fiber modules (or a combination of the two). The connection is between the Mikrotik and an Intel 82599 adapter. The connection will establish and then either immediately stat an never ending up/down cycle or sometimes will appear stable for several seconds only to fail.

I have bypassed the CRS310-1g-5s-4s+ by using an different switch and the connection is fine with any of the SFP+ modules or a combination of. I have tried turning off auto negotiation on both sides and a combination of on/off on one side or the other.

Note that I have 10Gtek DAC connections to this switch that are working ok.

The logs on the CRS310-1g-5s-4s+ only show the connection going up/down rapidly and repeatedly.

I contacted support over 4 days ago but have not heard anything back (honestly leaves a bad taste in my mouth). I do see some posts here indicating that other people have had issues with CRS series devices and fiber modules. Anyone have any suggestions, or am I at the mercy of Mikrotik providing a fix?

Tnx.

Why is this post not public? What's the point of this forum if you're not allowed to ask for help?

Well, I switched the 82599 adapter for an HP NC552SFP and it's working swimmingly. Still no word from support, which is disconcerting and the fact that it took two days for this post to show publicly here is sad.

I found 7.4.1 to be the most stable for the CRS300's, although my 310's ran fine on 7.6 too (with L3HW offload).

The question I have is whether you also upgraded the firmware at 7.4.1 (or higher).

It also seems you found an answer/resolution. It may not be a MikroTik specific thing, but a driver or hardware issue with the Intel card.

I had inconsistencies with some Cisco switches and Motorola wireless gear that required hard-coding the ports to the desired speeds (i.e. disabling auto-negotiation). You can safely assume there was no vendor fix from either party for that problem.