Hello!
Yesterday deployed three new CRS125-24G-1S and all have very strange bug:
All PCs connected to switch temporary loses connection to local network resources. Looking at Unicast FDB table of my CRS and i see the 1023 items on ether1 (uplink port). This count not growing, stay the same for hours.
Downgraded to 6.11 and i see that Unicast FDB table contains more than 1200 items for ether1, bug disappeared.
When this limitation was added and why?
All switches at 6.28 firmware (update to 6.29 doesn’t help me), all at default configurations, only 3 tagged vlans added and NTP adjusted.
CRS has a limit 1023 MAC adresses per port.
Its normal to a switch to have a mac address limit
Do you have 1023 MAC in your topology???
Can be an attack
Yes, i have over 1023 MAC’s in topology, they splitted to some VLAN’s, what’s wrong?
I cannot see this limitation on ROS 6.11, when it was introduced and why? I’ve found nothing in changelog about this.
Simple DES-12xx series working well in place of CRS, what i’m doing wrong?
well 1024 MAC thats a /22 ip network on a single port thats a lot of hosts for a single port, even on service provider networks.
the interesting point if why the problem is happening now
There are new CRS switch improvements in RouterOS v6.32rc5 which allow to overcome 1023 MAC learning limit per port.
Here are command examples to unset it on CRS ether1 port, similarly they apply to other ports
/interface ethernet switch port unset ether1 learn-limit
or
/interface ethernet switch port set ether1 !learn-limit