* fix(region): host health use sync map
Signed-off-by: wanyaoqi <d3lx.yq@gmail.com>
* fix auto migrate on host shutdown
Signed-off-by: wanyaoqi <d3lx.yq@gmail.com>
support usb controller type: qemu-xhci, usb-ehci.
qemu-xhci is much more virtualization-friendly when compared to EHCI and UHCI.
but guest supports XHCI which should be the case for any operating system
released around 2010 or later.
So sometimes wo need use usb controller type EHCI.
Signed-off-by: wanyaoqi <d3lx.yq@gmail.com>
If disk driver not assign, default use previous disk driver.
Bus windows server can't recognize hot attached 'ide' disk,
use 'virtio' replace 'ide' driver on diskattach.
Signed-off-by: wanyaoqi <d3lx.yq@gmail.com>
- host: host health reconnect etcd add timeout
- region: request health status use 'GET' method
- region: only migrate running guests on host down
- etcd: keepalive use background context
Signed-off-by: wanyaoqi <wanyaoqi@yunion.cn>
Before this patch, host guests will not migrate on host manual shutdown.
This patch from k8s node status check host still on line, if node status
not ready, region will migrate guests on host.
Signed-off-by: wanyaoqi <d3lx.yq@gmail.com>
* fix(region,host): auto migrate on host down
This patch fix source servers not shutdown on network unreachable.
Restart etcd session add timeout, and shutdown servers add force
option incase host stuck on scriptStop.
Signed-off-by: wanyaoqi <wanyaoqi@yunion.cn>
* host_health: add check network available
On host etcd keep alive failed, add check networks is available.
If networks available, may be is the etcd not work, in this case
we will net shutdowns servers and wait for etcd ready.
Signed-off-by: wanyaoqi <d3lx.yq@gmail.com>
Guest network add attribute num_queues, one is default.
And we can assign queues on guest create, eg:
server-create --net num-queues=4
Signed-off-by: wanyaoqi <d3lx.yq@gmail.com>
Co-authored-by: wanyaoqi <wanyaoqi@baidu.com>
