If you don't want your Nexus 5600 or Nexus 6001 switches to just … … … freeze, you'll need to replace their BIOSes.
So says a new Cisco field notice that explains “might experience hangs or reboots during a storm of non-maskable hardware interrupts (NMIs) to the main CPU.”
This one's a Cisco SNAFU: the company admits it is caused by “non-essential ports on the PCIe bus transmitting data” which means the poor old CPU is being it with so many messages it can't ignore that it eventually just falls over.
That sound you hear? Someone in the PCIe bus expert team at Cisco drawing breath very, very sharply.
The only way around this mess is an upgrade to the BIOS, which can be accomplished with an upgrade to NX-OS. Versions NX-OS 7.0(8)N1(1) and 7.3(0)N1(1) have the fix already. If you're on NX-OS version 7.1(4)N1(1), keep an eye on Cisco's various emissions as your upgrade is coming soon.
Cisco prides itself on quality - and today put out a blog post titled "Getting Quality Control Right the First Time" in which it claims to be very good at helping other organisations control quality. So mistakes like this aren't supposed to happen.
Having said that, the Borg has made a few howlers of late, like forgetting to ship management software with security appliances and shipping devices with the wrong default passwords. This one's a less obvious fault to have let through, but still far from a good look. ®