Community discussions

MikroTik App
 
stoser
Member Candidate
Member Candidate
Topic Author
Posts: 123
Joined: Sun Aug 21, 2016 12:04 am

Strange Behavior on Bridge with bad link on 1 port

Wed Sep 28, 2022 6:18 am

I have a Hex POE (RB960PGS) configured as a Bridge on all five ports plus the SFP port. FW 6.48.6. Hardware offloading is enabled on all ports. The SFP port is locked to 1Gbps. All other ports have auto-negotiation enabled. Ethernet Ports have port isolation enabled to all other ethernet ports, so Ethernet ports do not forward traffic to each other. They only forward to the SFP. No other configurations. I have many other routers of the same model at many other locations performing well with the same config.

One of the ethernet ports is misbehaving, currently auto-negotiates to 10Mbps, and often loses link. Normally it auto-negotiates at 1 Gbps. This is an installation at a tower in a remote location, co-located with FM radio transmitters. I suspect that the cable is suffering from RF interference. I will travel there to check in a couple of days. Until then I am trying to write a prioritized list of tasks and equipment so that I can use my time efficiently when I go there.

PROBLEM: The strange behavior is that when the bad 10Mbps link starts to get saturated (at about 8.5 Mbps) the other ports on the bridge start to slow down to about 10Mbps, and the latency to/from the devices connected to these ports goes from about 15ms to approximately 250ms. Their negotiation speed does not change, but they begin to behave as if they were on 10Mbps, but in an unstable way. CPU and memory usage does not increase. If I disable the faulty port, the other ports speed up again and the latency goes back down to 15ms. Also, sometimes if I disable the faulty port, the others stop trafficking for 10-20 seconds while they re-negotiate their link speeds. Yet other times, when I disable the faulty port the watchdog forces a reboot, indicating an irrecoverable crash or hardware issue.


QUESTIONS:
Is this normal behavior? Does a faulty link on one port affect all of the others on the same bridge? Or am I observing a hardware issue and should replace the unit?
 
mkx
Forum Guru
Forum Guru
Posts: 8936
Joined: Thu Mar 03, 2016 10:23 pm

Re: Strange Behavior on Bridge with bad link on 1 port

Wed Sep 28, 2022 9:22 am

TL;DR: I feel your pain.

The effect you're seeing is probably result of buffering in switch chip ... and I suspect that switch buffer is global, i.e. same for all switch ports. And when it gets full, switch can't perform the first part of "store&forward" stance which seems to be the most often mode of operation of switch chips.

It is possible, that speed down-negotiation is caused by hardware failure, but you'll have to check it yourself. In scenarios you're describing using shielded cables (STP) would quite probably help. I'm not sure if RJ45 ports on hEX PoE have grounded cages, so it might be necessary to use other means of grounding the shield (on one end only!).
 
stoser
Member Candidate
Member Candidate
Topic Author
Posts: 123
Joined: Sun Aug 21, 2016 12:04 am

Re: Strange Behavior on Bridge with bad link on 1 port

Wed Sep 28, 2022 2:51 pm

Good morning mkx.

1) I considered the buffer overrun, it just seemed strange that the buffer would be global/shared across all ports... As a last resort I will dedicate a dumb switch to the faulty device to mitigate the effects on the other devices. At least that way they won't share a buffer.

2) The Hex POE does not have a grounding lug but does have metal cages for ports, so I assume that they are connected to ground on the board, but I have never taken one apart to check. The site is running on a 24V battery bank, directly connected to the Hex POE, with only a fuse in between. The battery bank Negative is grounded, and I am using FTP cable for all antenna connections (they don't sell STP with a mesh where I am located, so foil is my only choice). I must add that the tower is made of steel, but there is no grounding system on the tower itself. There is only a grounding system in the shack. I always thought that the FTP would be grounded via the Battery bank that is grounded, but now I am having doubts. I will check voltage between ftp drain wire and battery negative to see if there is a difference, and connect the drain wire directly to grounding point if there is a difference.

3) If none of this works I may have to install FO ... The only problem is that I would have to run DC up to the tower, and theft of any cables thicker than 4mm2 is very common...

If anyone has any other thoughts please share. I am heading up to the tower/mountain today. Thanks in advance -

Who is online

Users browsing this forum: cdemers and 7 guests