ROS 2.9.38 and BIOS 2.6 - Ethernet issue

Hardware MikroTik hardware general
1

Gday,

I have recently upgraded a RB532 and a RB112 to ROS 2.9.38 and saw there was a updated BIOS (2.6) available. After the upgrade both RB’s are displaying the same problem. The ethernet has gone very unstable, causing many dropped packets…it seems to be affecting large packets a lot more than small ones. The switchport’s connected to the two upgraded RB’s had 1000’s of “alignment errors” and 10000’s of “collisions”.

Thru trial and error i have worked out that this problem can be minimised by switching ether1 auto negotiation off and switching speed to 10Mbit, half duplex.

Both boards were fine until i upgraded to 2.9.38 and BIOS 2.6.

They connect to a managed switch and locking the rate to 10Mb half duplex on the switchport also had the same effect as the ROS setting.

I have been unable to find a changelog for the 2.6 BIOS.

Has anyone else seen this problem?


Cheers…ƒil

Edit:- I upgraded another RB112 and don’t see the same problem. This one connects directly to an intel pro 100 interface in a laptop…seems to be fine. May only affect auto MDI/MDIX switches.

2

You found the solution on your own… :slight_smile:

RBs have auto MDIX ports, which have a strong tendency to cause trouble when connected to other MDIX ports. I have seen this happening with some Cable and DSL modems, as well as some cheap NICs.

With the RB5xx series you can disable Auto MDIX on Ether1, not sure if you can do it with the RB1xx series.

I don’t remember the exact command, but you should be able to find it in the manual, it’s a setting which you can only change through terminal (no GUI checkbox for this one)

3

for auto mdix to actually work you can’t be using crossover cables. but still there are some chipsets that just doesn’t want to talk to each other.

4

I have replaced the switch with a Cisco 2924 (non-MDIX) and re-ran new cat-5 cables. Cable length is around 15M, using good quality Cat-5. 48V POE.

The RB112 doesn’t seem to have the problem after I replaced the switch but the RB532A still does. It has remained consistant thru 2.9.39 and now 2.9.40.

It happens every time I soft reboot, either from console or winbox. Once I power cycle ether1 connects instantly. Soft rebooting several times doesn’t fix it either.

After a soft reboot (with no connectivity) the ether1 status is showing as “no-link” but the link is negotiated at 100Mb, full duplex.

Rebooting the cisco 2924 has no effect on the problem.

Do you think that this particular RB532A is a lemon?

5

Are you talking about CAT5 or CAT5e?
What’s the manufacturer and product code?
Which color code are you using?

And i’ve noticed that some units have trouble if the ethernet interface is reset and you are not using auto speed/duplex. (Which is a driver issue).

Have you ensured that the jumpers are set correctly for 48volt PoE?

Does it work better if you set the RB to 10mbit/half duplex?