Maintenance mode
Before you reboot a node, update its BIOS or change its hardware, put it in maintenance mode. Its VMs move to other nodes first, so they keep running while the node is down.
Maintenance mode needs at least two nodes, and another node online to take the VMs. The Maintenance button shows only while the node is online.
Start maintenance
- On Nodes, click Maintenance on the node's card.
- Click Move its VMs off.
The VMs move one at a time, each to the node that holds its copy. That only sends the latest changes, so each VM stops for about 20 seconds. A VM with no copy, or whose copy's node is offline or in maintenance itself, goes to the online node with the most free memory; it is copied there while it keeps running, so it too stops only for the last changes. The card shows Moving its VMs to other nodes with the VM being moved and a count.
When it is done, the card shows Maintenance, says how many of its VMs run on other nodes until maintenance ends (or It runs no VMs.), and It is safe to reboot or shut it down.
While in maintenance, the node takes no new VMs and no moves, and is not chosen to hold a VM's copy.
End maintenance
- Bring the node back online.
- Click End maintenance and move them back on its card (End maintenance if it moved none).
The same VMs move back, one at a time; the card shows Moving its VMs back. A VM deleted in the meantime, or already back, is skipped. Maintenance survives a reboot: the node remembers which VMs it moved off.
If a VM can't be moved back, the node stays in maintenance and its card lists the VM with the reason. Click End maintenance and move them back again to retry the ones still elsewhere.
When a VM can't be moved
If a VM can't be moved, the card lists it with the reason, and the node is not reported as safe to reboot. This also raises the Maintenance could not move VMs alarm (see Alarms). Fix the cause, or move or stop the VM yourself, before rebooting.
Maintenance and retiring a node
Retiring a node starts with the same maintenance step.