[Ksummit-2012-discuss] [ATTEND] writeback and kernel testing
fengguang.wu at intel.com
Sun Jun 17 01:50:12 UTC 2012
On Sat, Jun 16, 2012 at 06:18:58PM -0700, Guenter Roeck wrote:
> On Sat, Jun 16, 2012 at 10:50:41PM +0800, Fengguang Wu wrote:
> [ ... ]
> > Yeah, that should be a pretty common impression. The hardware
> > capability and the fruit of catching one bug per day turns out to be
> > much higher than my expectation :-)
> How many servers do you actually need to compile 25k kernels per day ?
> With my little system (i7-2600, SSD) I get about 1,000 randconfig kernel builds
> per 24 hours, so it should not be too many.
Two Sandybridge servers, totally 32 CPU cores.
> Imagine an extended patchwork where patches are not only published, but also
> automatically merged into the kernel and built. Thinking about it, and given the
> number of patches submitted each day, that should actually already be possible
> with just a single server.
Yup. Roughly speaking, one Linux release has about 10000 commits,
dividing that by 50 days, we get 200 commits per day. That made me
think that "commit-by-commit compilations may be feasible".
Currently ~400 commits are build tested per day, each does ~30
configs. In theory I could double the number of randconfigs w/o
overloading the system. However I'd like to keep good respond time in
rush hours. I think Randy runs more complete randconfig test sets on
linux-next anyway, so this is not a big concern.
I'd expect the runtime tests to eat more resources. One possible
improvement is to adaptively start kvm based runtime tests at the
compile servers' idle time. Also the servers could be better utilized
to run some extra heavy tests against each branch head during the
More information about the Ksummit-2012-discuss