Am 08.09.2017 um 09:55 schrieb Christian Ehrhardt <[email protected]>:
> I happened to track down a similar issue in 2.10 in qemu, but I checked and > the issue I found was only applied after 2.8. > Also I didn't have the I/O errors like you have above. > > You might check if you happen to have real I/O errors (showing up in demsg) > or if these are only qemu internal on the blockjob that the migration > triggered. I do not see any I/O errors in log or something showing up in dmesg. When I start the migration, VM keeps running on source host. On the target host, the qemu process for the vm appears in "virsh list" in "paused" state. Then both qemu processes stay in this deadlocked situation forever. However, I can see (of course encrypted) network traffic between source and target system on libvirt tls port 16514/tcp. I think this could be some kind of polling request from the target qemu instance, as the source process doesn't start the data transport (?). However, I don't have any idea how to debug this any deeper. -- Michael Schulz
