We may have found some regression between 6.36.4 and 6.37.1+ when running OSPF/LDP/VPLS between a CCR1009 and Juniper MX series router.
What we see when running 6.37.1+ ( tested all 6.37 and 6.38 releases )-
- Some MAC addresses have their traffic dropped on the floor, traffic coming back from the Juniper side is never delivered to the device behind the CCR, we can confirm that this traffic is valid and has the proper MPLS information attached to it. It didn’t drop ALL traffic but rather about 90%.
- We discovered this happening at two different physical locations, both with different underlying telcos providing L2 Metro Ethernet back to our core ( Juniper side ).
- After attempting to upgrade both sides with no joy, on one side we decided to swap in a CCR1009 pre-loaded with 6.36.4, we did not attempt to upgrade but rather it was a hail mary to see if it acted differently, once we swapped it in the affected MACs/devices were no longer having issues.
- We then decided to try and downgrade the other CCR to see if it was a hardware issue or a ROS issue, downgrading the other device resulted in the same success therefor we have concluded this to be some really weird regression… has anyone else seen this?
I’m trying to replicate this in a lab so we can isolate it further but I’m hoping maybe someone at Mikrotik has seen this or it has been fixed in 6.39 or something, thanks guys!