events/0 got you down?

I use CentOS 6.6 as my desktop with vmware player running VMs and I recently added an esata card to connect some extra external storage.  Over the last week, as I was running experiments the consisted of heavy cpu and I/O, at some point the performance of the machine would collapse and everything including the desktop slowed down.  The issue persisted, albeit with some relief, even after shut down all applications.  Looking at top, I saw events/0 using 99% cpu.  So, I had my culprit.

Google had few results and most of them were from 2010 and 2011 but they did point me in the right direction.  It appears that the events processes handle interrupts.  Lucky for me, I run munin on my machine, so all I had to do is look at the interrupt graph.  At the exact time of the slowdown the nouveau (the open source Nvidia driver) device considerably spiked.  Through Google, I wasn't able to find any other people having this exact problem, but after reading many a few posts, two themes emerged.  Either the PCIe cards are somehow sharing the same interrupts (my video and esata cards) or there's a bug in a device driver, in this case, nouveau.

Unfortunately, because it's for work and I needed to get things moving, I couldn't be too scientific about it, so I did two things at once.  I finally installed NVidia's proprietary Linux drivers and moved the esata card to a different slot.  It all works fine now.

John