Urban Boquist wrote:
28 jun 2009 kl. 15.39 skrev Jean-Yves Migeon:Does it happen only during domU start, or also during normal operation, like when DHCP requests new leases?All the domU:s are netbooting currently - so DHCP is only at boot time. But I'll try to set up a new one and check that more carefully.
Thanks; I'd like to see if it is an initialization issue, or some routing/handling problem between bridge and xennet IO rings.
My feeling is that it might happen now and then also during normal operation (I get NFS server not responding messages) but the situation gets much worse when a new domU is started. Also, during the time a new domU is created communication with the other domU:s seems to be affected also. They "freeze" for say half a minute. It's almost like the entire bridge interface "stops" for a while.
From my PoV, backend devices having no thread context, one xennet could thrash the rest of the xennet communications until situation calms down a bit.
It would be interesting to observe if the "freeze" concerns only xennet communications, or if console/vbd/domain scheduling are affected too.
-- Jean-Yves Migeon jeanyves.migeon%free.fr@localhost