[Qemu-devel] Re: [BUG] Regression of exec migration

[Chris Lalancette]

Pierre Riteau wrote:
> Commit 907500095851230a480b14bc852c4e49d32cb16d makes exec migration  
> much slower than before.
> I'm running the latest HEAD of qemu, on Debian Lenny 5.0.2.
> 
> I'm migrating a fully booted Linux VM (also running Lenny) with 128MB  
> of RAM to a file, using the following command: migrate "exec: cat >  
> vmimage". The resulting file has a size of 57MB (because we save only  
> what is allocated from the 128MB).
> With the current HEAD, it takes from 15 to 40 seconds (it's variable)  
> to perform the migration to the file.
> With commit 907500095851230a480b14bc852c4e49d32cb16d reverted (or just  
> commenting the "socket_set_nonblock(s->fd);" statement), it takes  
> about 3 seconds.
> 
> I did a little debugging and it appears that, when in non-blocking  
> mode, a write to the popen'ed process which fails with EAGAIN (because  
> cat is busy writing data on disk) will only be retried after 100 ms  
> (the next timer tick?).
> 
> Attached are two traces of qemu's output with DEBUG_BUFFERED_FILE  
> defined.
> I modified the dprintf macro to print the result of  
> qemu_get_clock(rt_clock) before the message.
> nonblock.log is with an unmodified HEAD while block.log has the  
> "socket_set_nonblock(s->fd);" statement commented.
> Notice the "backend not ready, freezing" messages when nonblock is used.

Hm, interesting, thanks for the report.  Let me try out a few things and get
back to you.

-- 
Chris Lalancette