* Li, Liang Z (liang.z...@intel.com) wrote: > > * Li, Liang Z (liang.z...@intel.com) wrote: > > > > > > First explanation, why I think this don't fix the full problem. > > > > > > Whith this patch, we fix the problem where we have a dirty block > > > > > > layer but basically nothing dirtying the memory on the guest (we > > > > > > are moving the 20 seconds from max_downtime for the blocklayer > > > > > > flush), to 20 seconds until we have decided that the amount of > > > > > > dirty memory is small enough to be transferred during > > > > > > max_downtime. But it is still going to take 20 seconds to flush > > > > > > the block layer, and during that 20 seconds, the amount of memory > > that can be dirty is HUGE. > > > > > > > > > > It's true. > > > > > > > > What kind of cache is it actually that takes 20s to flush here? > > > > > > > > > > I run a script in the guest which do a dd operation, like this: > > > > > > #!/bin/sh > > > for i in {1..1000000} > > > do > > > time dd if=/dev/zero of=/time.bdf bs=4k count=200000 > > > rm /time.bdf > > > done > > > > > > It's an extreme case. > > > > With what qemu options for the device, and what was your device backed by? > > Very simple: > ./qemu-system-x86_64 -enable-kvm -smp 4 -m 4096 -net none rhel6u5.img > -monitor stdio > > And it's a local migration. I will do the test between two physical machines > later.
OK, but for shared storage you would have to add cache=none (or something like that), so that would change the behaviour anyway. Dave > > > Liang -- Dr. David Alan Gilbert / dgilb...@redhat.com / Manchester, UK