Hi Mike,
This strikes me as more of a vdsm problem than ovirt-node directly. Once node is registered to engine, we hand over all control of libvirt and networking (among other things) to the engine to manage.
I just tried migrating between 2 nodes but that also fails.
Doing a connect between the 2 nodes from a virsh I get the same connection refused I get when trying to connect the node from the vdsm on the engine server. So they can't do qemu+tls:/ipaddress/system to each other. Stranger still I just tried the qemu+tls connect to the vdsm which worked before, now it's also a connection refused I get back.
If migration is failing due to some setting on the node, then vdsm should probably be changing that setting when it comes online.
I'll attache a grep from the vdsm.log while the migration was started from nodes. Maybe someone can take a look and hopefully find something.
Time difference between the servers is about 40 seconds.
Mike
Michel
vdsm-devel@lists.fedorahosted.org