2867be: LAN ports dropping when most in use.

Folks

I don't know if this is just my unit, but after spending more time on this than strictly necessary, I've found that if I run all the LAN ports with both Eth WAN ports, the router behaves very weirdly, most obviously with missing ping responses on wired the LAN ports, difficulty autonegotiating port speed, and trouble obtaining DHCP addresses.

I'm running the WAN P1 port a 1Gbps and WAN P3 port at 10Gbps. It's especially bad with an SFP+ nodule in the LAN P4 port, but is also unreliable with no SFP+ module and all of the other four LAN ports populated. Frankly, it's unusable.

My workaround is to use a separate 2.5Gbps switch hanging off the LAN P5 (2.5Gbps) port.

This is with the unit powered by the supplied AC adapter.

I would try dropping the WAN P3 port down to 2.5Gbps, but I have no access to the ONT, and, for some reason, the P3 port only allows Auto-negotiation.: there is no option on this particular port to pin a port speed, unlike the others.

Firmware is 5.4.4.2

Has anyone else encountered this?

Just checking here - you have the right settings in the interface to be using ports P3 and P4 and NOT P2.

Out of interest why buy a 2867 and not a 2928 if you're not using the DSL port anyway ? Or are you, and you've just not mentioned it above?

Thank you for the quick response.

To answer you last question first, regarding DSL. This is risk mitigation. The plan has always been to do a lift and shift first, replacing the existing 2865ac with the 2867be, using the existing DSL and 1Gbps FTTP. Once that got working, with all the stuff like VPN, DDNS, failover etc, then add a new 2.3Gbps FTTP, removing and decommissioning the DSL connection.

> Just checking here - you have the right settings in the interface to be using ports P3 and P4 and NOT P2.

The interfaces in use are...

P1 WAN1 1Gbps FTTP 900Mbps service to ONT1
P2 WAN2 DSL (unconnected and disabled if WAN3 is used)
P3 WAN3 10Gbps FTTP 2.3Gbps service to ONT2 (This port only supports auto-negotiation, which negotiates to 10Gbps with the ONT)
P4 LAN SFP+ copper to 2.5Gbps switch
P5 LAN 2.5Gbps to 2.5Gps switch
P6 LAN 1Gbps to 100Mbps device
P7 LAN 1Gbps to 100Mbps device
P8 LAN 1Gbps to 1Gbps device

If I leave the WAN ports alone, but remove LAN ports P4, P6, P7 and P8, placing them into a separate switch running off P5 instead, everything works fine.

Whether the router's WiFi is on or off makes no difference.

What has taken time to diagnose is that most of the testing was done on the bench with only a single PC plugged into one LAN port, with two WAN ports running (either WAN1 1Gbps plus DSL, or WAN1 1Gbps plus WAN3 10Gbps). Of course everything works fine on the bench, including load balancing, VPN etc etc. But when placed in situ, with the remaining LAN ports populated, things start going awry, with ports dropping, pings going unanswered sporadically but often, and the unit struggling to boot.

OK all makes sense.

It doesn't help - but - I have a 2928 with just P3 & P4 in use (pretty much as per yours there).
P4 uplinks to a switch so P5-8 are not in use.

Is there potentially anything odd about the 4 devices you have plugged in - have you tried cycling through using only three at a time to see if there's a odd one that causes the issues to start ?


Just noticed - P4 & P5 to the same switch ?  Potential loop ?

Hello Ian

Thanks again for your input.

I suspect it may be some kind of power limitation or thermal issue internally. I did try gradually taking devices in and out, and it did improve things, however I never found a configuration that seemed to make it stable without just putting everything into an external switch.

Unfortunately I've had to pull it out of production and revert to the old 2865 for another reason: the Bind-IP-to-MAC that I imported to the 2867be from the 2865 stopped working. I had unit tested this before putting the unit into production, but for some reason it's decided to stop working. It serves up DHCP addresses fine, just not the ones in the bind-ip-to-mac list!

The bind-ip-to-mac issue was resolved by (1) cleaning up the exported .cfg file from the 2865 by replacing spaces with dashes in names in Notepad++ with some reg-exp foo, and (2) once imported into the 2867, export from the 2867 then re-import.

I don't know if (1) is required, but I'd done it anyway after being unable to get this to work. Also note that the new backup format isn't in text any more. (2) was recommended by Draytek phone support.

I also recommend shutting down as many devices as you can that received the wrong DHCP addresses before bringing up the network again to avoid any DHCP conflicts.