Everything the panel does on a node is a task: create, power, reinstall, snapshot, backup, move. This page covers the server's page, its power actions, how servers come back after a node reboots, and how to follow and stop tasks.
The server page#
Open Servers and click a server. The page has a header, banners, and tabs.
Header, on every tab: the name, status, address (with Copy), system, location (administrators see the node), size, the power buttons Start, Restart, Shut down and Force power off, and Open console (see Console). Nothing in the header destroys anything.
Banners appear above the tabs when they apply: being built or reinstalled (with progress), being deleted, in error, a task under way or the last one that failed (with Open the task that failed), Waiting for the node, Under way (a long task you may stop), suspended, an upgrade the node has not confirmed, the node not answering, DNS records not created yet, planned maintenance of the node.
Tabs (a tab shows only when the reader may use it and the server has it; the open tab is in the address as ?tab=name):
| Tab | What is there |
|---|---|
| Overview | Connect (SSH command, root password; the rescue system while in rescue; the installer for an ISO install), notes (administrators and resellers), usage now, guest facts, and Details (ID, addresses, plan, sizes, snapshots, node, owner, build log, placement). |
| Network | Addresses, reverse DNS, port forwards, private networks, Reconfigure network. See Networking. |
| Firewall | The server's firewall. See Firewall. |
| Storage | Disks, ISO drives, boot order and boot once. See Disks, CPU and GPU and ISOs. |
| Snapshots | See Snapshots. |
| Backups | See Backups. |
| Usage | CPU, memory, disk and traffic charts, speed limits. |
| System log | The last lines the server's console printed, fetched from the node on opening and on Refresh. |
| Settings | Recovery (rescue, reset password, reconfigure network), network card, disk driver, advanced options, the screen console switch. |
| Manage | Upgrade, change owner, move, high availability, limits, snapshot policy, clone, start on boot, suspend; then the Danger zone: Reinstall and Delete server. |
| Tasks | The server's last tasks. Administrators see types and can open each task; others read them in plain words. |
| Activity | Administrators and resellers: the server's last 25 audit entries in words. |
Containers have no Storage, Snapshots/Backups (on older agents), System log or Settings tab; their reset password and SSH keys are on Manage.
Power actions#
| Button | What the node does |
|---|---|
| Start | Starts a stopped server. |
| Restart | Restarts the server. |
| Shut down | Asks the guest to shut down (ACPI). A guest that ignores it keeps running. |
| Force power off | Cuts the power at once, like pulling the plug. Unsaved data in the guest is lost. |
The buttons are disabled while another task of the server runs, while it is suspended, and while its node does not answer. Customers need the power abilities of their plan ("power", "power off"). In rescue, power actions act on the rescue system.
When a node does not answer, its servers are locked, not hidden: the page shows the last known state with a muted light and "Node … unreachable since 14:05 UTC; the server may still be running." Tasks asked meanwhile wait under Waiting for the node and are sent when it answers; whoever started one, and administrators, can Cancel it there.
Start on boot#
Every server has Start on boot, on by default. After its node reboots, the node starts the servers that have it and were running when the node went down:
- oldest first, one at a time, waiting until each runs (at most 60 seconds), then
boot_start_intervalseconds (default 10) before the next; - a server its owner shut down stays off; a suspended server is started paused; a server in rescue starts its rescue system;
- a server that fails to start is tried once more at the end; one that still fails is named on the node page.
To change it for a server: on the Manage tab, turn the Start on boot switch. The owner, its reseller and administrators can; on a suspended server only administrators. The setting goes to the node at once. If the node keeps another value after 5 tries, the switch says so; turn it again to resend.
The node page shows "Starting servers after boot: 3 of 12" while the start runs, and for a day after, how it went.
To change the pace on a node, set the interval in /etc/vmstack/agent.yaml (0 to 600 seconds) and restart the agent:
boot_start_interval: 20
systemctl restart vmstack-agent
Note: A server counts as "stopped by its owner" only when the guest was shut down cleanly while the node was not shutting down. A crashed or killed server is started after the next boot. A script that shuts servers down cleanly more than about a minute before a node reboot makes them count as stopped by their owners.
Requires agent 0.21.4 or later. On an older agent the switch is saved and says it applies once the node is updated. API: PUT /servers/{id}/start-on-boot with {"start_on_boot": true|false}.
Tasks#
Find a task#
- System › Tasks (administrators): every task, newest first, with Status, Type, Node, Started by, Started and Duration. Filter by status (Queued, Running, Succeeded, Failed, Cancelled, Unknown), node and type.
- A server's Tasks tab: that server's last tasks.
- A task's page: progress, the error and what the node said, details (node, who started it, times, time limit), the result (passwords, tokens and keys are never shown), and links to the build log and the placement record where there is one.
A task is Unknown when the node stopped answering while it ran; its real outcome is fetched when the node is back. Nothing is ever sent twice.
Cancel a queued task#
A task not yet sent to the node can be cancelled by whoever started it, or an administrator: Cancel under Waiting for the node on the server page, or on the Tasks page. It is never sent, and what it was about goes back to how it was (a cancelled create removes the server and frees its addresses).
Stop a running task#
Long tasks can be stopped while they run, with the node cleaning up what they started. The node needs agent 0.44.0 or later; on an older agent the Tasks page shows "Agent too old" instead of a button.
- On System › Tasks click Cancel on the running row, or Cancel task on the task's page. Customers (and resellers) get Cancel in the server page's Under way panel for their own server's backups and restores, when their plan lets them take or restore backups.
- Read what stopping does, then click Stop the task (or Keep it running).
The task shows Stopping… until the node has cleaned up, then Cancelled. If the work finished first, it ends succeeded (the cancel came too late).
| Task | What stopping does |
|---|---|
Backup (backup.create) |
The backup stops, what it wrote to the destination is deleted, and the server keeps running on its own disks. It counts as a failed backup, shown as Cancelled. |
Restore (backup.restore) |
Stops before it touches the server's disks; the server stays as it is. Once the node swaps the disks it can no longer be stopped. |
Backup check (backup.verify) |
The check stops. |
Image download (cache.fetch), ISO insert (vm.iso) |
The download stops and its partial file is removed. |
Make an image (vm.image.capture) |
Stops; the unfinished image is removed. The server stays shut down. |
Create or import (vm.create) |
The node removes what it made; the server is deleted from the panel and its addresses freed. For a Proxmox VE import, the VM in Proxmox VE is unchanged. |
Move (vm.migrate.export, vm.migrate.import) |
The move is taken back; see Moves. |
Other tasks cannot be stopped: reinstall, resize, disk moves, snapshot tasks, storage tasks and short tasks. A cancel refused because the task is in a step that must not be cut short says so ("Not stopped: …"); you can ask again later. If the node does not answer, the panel tries 5 times, then says "Its node did not answer, so it was not stopped. Try again."
API: POST /tasks/{id}/cancel. Webhook: task.cancelled.