I am experiencing an issue with v6.47.2:
As it looks ethernet link handling on RB3011 has problems.
What I did: Upgrade RB3011 from v6.47.1 to v6.47.2 What happened: RB3011 became unreachable What turned out:Trunk link between RB3011 (ether10) and CSS326 (port3, SwOS v2.11) became unreliable even if no RX/TX errors being reported Further actions: Play with Autonegotiation and Flow Control did not improve the link. But several (!) full power cycles of RB3011 and CSS326 changed something. The heck knows what!
Status now: Link somewhat works but link setup is unclear as different modes are reported (yes, I can’t believe that myself):
CSS326 webif: autonego on, 1G, full duplex, flow control on
RB3011 webif: autonego on, 1G, full duplex, flow (RX&TX) control on
RB3011 console: autonego on, 100M, full duplex, flow (RX&TX) control on
Take it as a warning and wait for an explanation first!
At least make sure you can powercycle your remote devices! (Not reboot but powercycle=plu-out-plug-in)
Actual link state in ROS is displayed using command /interface ethernet monitor ether10 once. With print you don’t really get actual interface state, you only get actual config … and speed=100Mbps in that printout is setting withich would have been applied if auto-negotiation was set to no (which it isn’t).
Any reason to limit advertised modes to only 100Mbps and 1Gbps full-duplex?
Make sure flow control settings match on both link partners.
That is then in line with the webif and makes perfect sense then. Thank you for that important hint.
So the conclusion boils down to a certain upgrade procedure to be performed for the RB3011
Upgrade packages
1a) Reboot
—> be sure to have direct access to one of the devices ethernet ports <—
Upgrade Routerboard (kernel)
2a) Reboot
Power cycle RB as reboot seems insufficient
Can’t say for sure, but looks like: Power cycle CSS as well
Otherwise your ethernet ports may not find together.
Or something even more weird happens, I just cannot imagine.
Predictable is certainly something else.