RB-1100 Port problems, and performance problems

Hardware MikroTik hardware general
1

Some questions for clarification. Have you tried turning auto off on ports 1-10 and forcing the connection to 100 half duplex? Does it work? Looks like something might be wrong making a full duplex link between that switch and the RB1100. Have any other equipment that is experiencing the same issues with the RB1100 ports 1-10?

The port 12 tests involving the 1800 are routed, correct? Do you think the half-duplex negotiations are causing these issues? Maybe the Nortel and the RB just are not going to play nice with each other. It would be interesting to see if you have the problems with any other piece of networking equipment.

As for the second set of tests, I’m assuming you’re routing, correct? What are the different speeds you’re seeing? And when you run the tests through the router, are the two routers being tested plugged into the same 5 port grouping? 1-5 and 6-10 each have one backplane to the CPU. You’ll want to have each router plugged into one group if you’re testing the maximum throughput. So, plug in one into port 1 and the other into port 6 and look at the performance. You should see an improvement compared to plugging the two units into port 1 and port 2.

RB1100 Port 11/12 issues
Can't connect when local user goes inactive
2

There definitely are big issues with the RB800/1100 ethernet. I’ve had an 1100 for almost a month now and cannot use it because of these problems.

My edge router is an old 1GHz Mini-ITX box with dual on-board NICs and a RB44. We’re pushing peaks of 72x12Mbps on our 100Mb F/D fiber-ethernet circuit. And the CPU gets up around 95%. I think the PCI bus on these small PCs just don’t have enough bandwidth to the CPU for packet routing.

Before I did anything, I first loaded v4.10 and upgraded the boot firmware to 2.27. Our uplink goes into a Cisco 3500 switch and AT&T has our port forced to 100Mb F/D (common auto-neg issues on Cisco gear). I have ether13 on the RB1100 forced to 100 F/D as well. The port randomly throws errors. I look at it and it’s at 100 H/D even though it’s configured for 100 full, auto is OFF. Plug back into the old router, works fine.

Ether11 and 12 randomly stop passing traffic for about 10 seconds and then the links go down and come back up, over and over. I had two Trango backhauls plugged into those ports. Works fine on the old router.

Ether1-10 didn’t seem to have link or negotiation issues or errors, but I haven’t tested for throughput problems on them either.

And today I upgraded to 5.0beta4, same problems all around. I sent MT support a supout from it after the testing. I hope they find the problem(s) and fix it.. soon.

3

Just a “me too” with RB1100 and Ports 11 through 13.

Does anyone have update on a fix? I need the bypass…

Thanks,

Andy

4

What is exact configuration for both end (auto-negotiation=yes/no, full-duplex=yes/no, speed)?

If one end has auto-negotiation enabled while disabled on the other, half duplex mode will be used on auto-negotiation end.

What results are you getting if you pass traffic through ether1-ether5, ether1-ether2 and ether6-ether7?

Please generate and send supout.rif file to support@mikrotik.com from configuration having these problems.

5

kewlkeed, poor speed could be result of duplex mismatch. We will investigate why this link does not work at forced 100FD.

verdonker, please send supout.rif file to support@mikrotik.com

6

I, have the same problem. we are waiting for a urgent solution.

Thanks

Luis

7

I was asked by kirshteins through the ticket I have open with MT support to back the CPU speed down from 1333 to 800 and the memory speed down from 533 to 400 and see if this problem still exists and send a supout. Did all that. Even backed both the CPU and memory clock down to 333MHz, no difference.

We have an F/O link from AT&T and the demarc/termination is a Cisco 3500 (or 3550, don’t remember) switch. They always hard set the copper ports to 100 FD. There are known auto-neg issues with Cisco gear and non-Cisco gear.

So this time, instead of driving all the way to the site and taking my whole network down, I tested on the bench.

Set ether13 forced to 100 FD. Set my laptop to the same. RX errors continuously. Tried two managed switches and a Cisco 1800 router, all forced to 100 FD, RX errors still.

Set ether13 to auto, laptop to auto, switches and router to auto. At first it didn’t work on my laptop. Unplugged the cable to bring the link down, plugged it back in, everything works fine.

So this tells me that, even though in Winbox the port is forced to a speed and duplex, it still auto-negotiates. With the other devices forced, the RB1100 still auto-negotiates, even though the interface status says it’s running full-duplex, it’s really only half-duplex while the other end is full. Duplex mis-match = errors and terrible speed.

Still don’t know why ether11 & 12 links suddenly and randomly drop out. I have every port set to auto-neg except ether13.

This is obviously some kind of ethernet hardware and/or software-to-hardware communication problem.

I needed this thing working like a month ago. Please fix it MT, please.

8

Did you test your Cisco Catalyst 3500 and RB1100 port 11-13 enable auto-negotiation and make gbit connection over twisted pair.

9

I experience the same issue. Connected an RB1100 Ether11 to a cisco ME3400 switch. Interface resets every 10 seconds or so. Ether12 and 13, same issue. All other interfaces works fine. Tried another RB1100 with the same issue. This is a really nasty error.

10

Is there are auto-negotiation enabled on both end or is it forced 100FD link? Please send supout.rif file to support@mikrotik.com

11

The interfaces are set to auto-negotiation. I probably won’t have time to replicate the issue as I’m leaving for holiday later today, but I’ll make sure to get back to you when I get back.

12

I see wrong interface statistics in Port 11-13.

Plugged in a device in one of those ports. Removed this device.
Now I see incoming Traffic >200MBit on this port.
Connecting s.th. to this port it looks ok. After reboot all is ok again.

13

Another “me, too”. Couldn’t get 100FD hard set on both sides to work between an RB110 and a Cisco 4507 on ports 11-13. Auto-negotiation had the port fail randomly.

Migrated configuration to ports 1-10 and all is well.

14

Another me too here. All 3 1100s have the exact same problems with these 3 ports. Errhhhmmmm… I REALLY HOPE I am wrong here but there maybe something more to this than just a simple software fix. It looks more like a hardware bug…

I have turned off auto negotiation everywhere and the problems seem to have stopped for the past 2-3 hrs anyway.

15

are you guys all restoring backups on these 1100’s to get them configured? I had a friend run into the exact same problem and I told him no way is it a hardware problem, although it sure seemed like one. I took control and ran a /system reset-configuration and reconfigured it from scratch and all those problems went away. I know a backup should have nothing to do with full duplex / auto negotiation, but something was seriously wrong to the point it wouldn’t work with a restored backup. After making it work I used diff to check the differences between the old config (before reset) and the new config - and there were no differences… so maybe it’s something that gets corrupted outside of the visible config.

Just a shot in the dark… same thing happens on the wireless config reset, for some reason it just fixes things.

16

Mine were new configurations (copy/paste a text template into the CLI), and applied straight to how the router shipped with a few of the default packages removed. The interface configuration in that template only deals with interface names, all other interface changes were made manually.

17

New config here also but speed and duplex negotiation doesn’t really have to do much with configuration in most cases when set to auto. This is really a hardware implementation problem and a pretty old one actually that used to happen quite a lot a few years back.

18

No, not a .backup file here. I exported the config from my old router, edited for the new hardware (removed ALL ethernet settings), and it imported IPs, routing, firewall, queues, etc. without issue.

5.0b4 or 4.10, doesn’t make a difference.

Set my laptop to 100 full, Rx errors on every port of the RB1100 forced to 100 full. Laptop and RB493AH works fine forced to 100 full. Laptop and managed switches, laptop and Cisco switches, laptop and Cisco routers all work fine. RB493AH to RB1100 ethers 1-10 works fine, 11-13 does not.

Anyway, I couldn’t wait any longer for the 1100 to be fixed. So I bought a Roc-Box from Tom @ Roc-Noc. It has seven Intel 82573L Gbit NICs and uses the PRO/1000 driver. Forced to 100 full works perfectly fine with everything. I have it running v4.10 right now on the edge of my network routing about 75Mbps and 12k PPS and the CPU only hits 10-15%. And it has an expansion slot for other NICs. I would’ve been perfectly happy with the RB1100 if it worked AND I wouldn’t have needed to install a new rack to fit the Roc-Box. But when MT figures out the 1100 issues, I’ll be happy to know that I have a backup router.

19

The workaround of this 1100 hardware glitch is to turn off auto negotiation. No more disconnects for me here after this.

THIS MUST BE DONE IN ALL DEVICES CONNECTED TO PORTS 11-13.

What I wonder though is if Mikrotik acknowledges the problem and is going to replace the boards having this hardware failure.

20

Try to test with recently released v5.0beta5 if possible.