CCR2004-1G-2XS-PCIe SFP+ modules crash host system

I’ve been very happy with my CCR2004-1G-2XS-PCIe for the past year. I upgraded to ROS 7.14 and suddenly my host crashed and would not post with a NMI error. After much troubleshooting, I discovered the host system will boot and work fine as long as nothing is plugged into the SFP+ interfaces. Next, I decided to try to upgrade to ROS 7.14.3 in case the issue has since been resolve in a newer release, only to discover that it fails to upgrade. Logs aren’t very helpful:

11:47:17 system,info router rebooted
11:47:17 system,error,critical kernel failure in previous boot
11:47:19 interface,info lo link up

Then I attempted to downgrade to the last version that worked. It fails as well with the same issue.

Any suggestions on where to go next with troubleshooting in recovering my router? I’m starting to think maybe something is wrong with the hardware.

I may have found the solution to my own problem.

http://forum.mikrotik.com/t/v7-15rc-testing-is-released/175351/1

*) sfp - improved system stability for CR2004-1G-2XS-PCIe (introduced in v7.14);

I’ll try 7.15rc tonight. Hopefully it gets promoted to full release soon.

Final update for the next poor unfortunate soul that find this due to the same issue.

7.15rc seems to have resolved my problem but getting this card to netinstall was more fiddly than I would have expected.

First, read and re-read the netinstall instructions. If you are running on a Windows machine, be sure to disable firewalls, run as Administrator, directly connect your computer to the management port on the router with nothing in between, and set static IPs for everything as indicated in the documentation.

Second, I never was able to get the card to netinstall using the reset button on the card. Instead, I used a Linux Live USB stick to boot into Linux and use the network interfaces on the PCI bus to connect to webfig and set the card to netinstall on the next boot (System → RouterBOARD → Settings → Boot Device → try-ethernet-once-then-nand. this setting is only effective on next boot and must be reset each time you attempt to netinstall again.

Once 7.15.rc2 was flashed to the system and router reboot initiated, my server immediately crashed with an NMI interrupt error. I disconnected the netinstall system and power cycled the server. Everything seems to have come back online correctly and 7.15r2 appears to have fixed all my problems.



Two things I would request on the next version of this router if possible:

First, it would be great if there were a way to power up the router outside a computer. Possibly header pins or a 12v Molex that accepts power input? I understand there may be issues with applying power on the header while also accepting power on from the bus, but if video cards can do it, surely there’s a way.

Second, while I’m asking… A USB header on the board would be a very versatile feature. It could be used in host mode to connect a USB storage device (either internally mounted or extended to for external connectivity on a separate PCI slot), or used in device mode as a serial console to a USB header on the motherboard.