virtio + raid1 = kernel panic

[john.newman]

Hi

Today one of my VMs started throwing kernel panics as I was trying to update it. I'm just wondering how exactly I can debug this problem to the root cause and work with whomever (?) is upstream to go about getting the fix merged in.

This is with 3.3.8 .. I am _trying to update world, and get over to 3.4 ... perhaps I will just update gentoo-sources out of band and see if the problem still persists there first. If I can even do that without getting the panic that is... Well, I could also backout this raid1 setup if that happens

It happens due to heavy disk IO on /var. This VM has a 'hybrid raid 1' setup for /var. It's a ramdisk + a physical partition, created with "write-behind" and "write-mostly". The idea is to have a full copy of /var loaded into RAM, such that mostly all reads will come from the ramdisk, and writes will lean there as well and eventually sink down to the persistent disk. On startup the array boots degraded just with the partition, then in localstart I add /dev/ram0 to it and it takes a second to rebuild. Really I think this is a pretty sweet idea, but I haven't gone about measuring the performance yet. I got the idea from here: https://bbs.archlinux.org/viewtopic.php?pid=493773#p493773 and here http://www.linuxquestions.org/questions/linux-general-1/using-tmpfs-is-not-improving-performance-761822/#post3719872 (Yes .. this box is plugged into a UPS )


I figured out how to get the serial console setup (very nice) and actually got the full trace: