Running the cluster
This is about daily operations. Creating VM instances is covered in kiss-configure.
Overview
Run kiss-overview
:
OK: kissclusterd.service is running
OK: Active target: primary-vms.target
VM status:
vm@vm1.service loaded active running VM vm1
vm@vm2.service loaded active running VM vm2
DRBD status:
0:vm1/0 Connected Primary/Secondary UpToDate/UpToDate
4:vm2/0 Connected Primary/Secondary UpToDate/UpToDate
5:vm2/1 Connected Primary/Secondary UpToDate/UpToDate
This shows the status of kissclusterd (which should be running, obviously), all defined VMs (that should be running if this is their primary host), and the status of the DRBD resources.
Starting/stopping VMs
VMs can be started and stopped via systemctl
, as
vm@NAME.service
. Stopping uses kiss-vm-monitor NAME powerdown
,
so it sends an ACPI shutdown signal and waits for the VM to cleanly
shut down.
Migrating VMs
VMs can be live-migrated via kiss-migrate
. This tries to catch any
errors, but be careful. kiss-migrate --all
can be used to migrate
all running VMs to the other node, e.g. if you want to shutdown one
node.
Taking a node offline, or rebooting it
If you just reboot or powerdown a node without migrating the VMs, they
will be shut down. Then, the other node's kissclusterd
will
eventually see this node is down and start them again. This should
bring an outage of about a minute.
To have a smooth reboot/powerdown, migrate them as shown above. When
the node comes up again, it will automatically start kissclusterd
,
both nodes will soon try to switch to primary-vms.target
, but the
other node will not shut down any VMs due to this, so this node will
not be able to start them.
Thus, you have to migrate them over (one by one) manually.