Craig Ulmer

Switch problems

2013-08-07 clusters net

Well, there's something you hope to not see when you're digging around in a network switch, trying to figure out why things aren't working. I've been having lots of problems lately with nodes in a cluster losing their nfs. It's probably an ipmi problem, but I thought I'd look into the 48port GigE switch to see if it could tell me more. It didn't, so I upgraded the firmware.

In the installation instructions they tell you to con-c the boot loader and type in 3 letters. This brought up the (hidden) developer menu that has additional options to replace the bootloader (in addition to the normal firmware). It also adds that Backdoor option. It's probably nothing but.. dunno. I think if I were in a Taiwanese company making switches, I would avoid using certain phrases in the firmware.

In the end, firmware worked, but nodes are still having network dropouts. Guess I'll be digging into ipmi again, soon.