Running VPS services¶
The admin panel¶
A Proxmox service's page in the admin area (Clients → a client → Services → the service) has the same panel the customer sees, plus the VM number and node, the pool address and any additional IPv4 addresses (with the network card each is on), an Open in Proxmox link and a field to link an existing machine. Staff may also work on a suspended service's server — to look into it, or to take a backup before it is terminated; a terminated or cancelled one is left alone.
The usual module actions sit above it: Suspend, Unsuspend, Terminate, Change Password, and Create Account while the service has no server yet.
Adding a VPS by hand¶
On a client's Services tab, Add Service:
- A new server — pick the product, the server and the order options (operating system, memory, cores…), tick Create the account on the server now, and PNLCS builds it at once exactly as for an order.
- A server that already runs — for a machine made before PNLCS or by hand: leave Create the account… unticked, pick the server, and choose the machine under Link to an existing account. The list shows what the token can see, and names machines another service already has. PNLCS tags the chosen machine for the new service and manages it from then on; one that belongs to another service is refused.
The same link is on an existing service's page (Link an existing guest, by VM number).
The form has no SSH key field: a server added by hand gets the customer's keys only when it is reinstalled from the panel with keys in the SSH public keys box.
What happens when¶
| Event | On Proxmox |
|---|---|
| Order paid (or placed / accepted, per Auto Setup) | KVM: clone of the template into the pool, with the next VM id in the range; tags and note; cores, memory, network, cloud-init (user, password, SSH keys, address, DNS, vendor snippet); disk grown to the product's size; protection on; started. LXC: the container is created from its template with the same settings and the SSH keys |
| Suspend (overdue invoice, or by hand) | Start at boot turned off, then shut down (forced after a minute) — a host reboot does not bring it back |
| Unsuspend | Start at boot on, started |
| Terminate | Stopped, protection lifted, deleted with its disks; its pool addresses, additional ones included, go back to the pool |
| Upgrade / downgrade | Additional IPv4 addresses are taken from the pool first; a pool that cannot give them all changes nothing. Then new cores and memory are set (a running KVM takes them at its next reboot); the disk grows if the new size is larger and is never shrunk |
| Every minute | The scheduler moves on reinstalls, restores, snapshot rollbacks and image downloads |
| Every hour | Disk usage and this month's traffic are read into the service |
PNLCS never deletes a machine that does not carry its mark for that service, and treats a machine that is already gone as removed rather than failing.
Traffic¶
Traffic is counted from Proxmox's own statistics for the current calendar month, so it survives reboots (the counters on the machine itself restart at every boot). It appears on the service as usage against the product's monthly traffic.
Capacity¶
Setup → Servers → Capacity on a Proxmox server shows, for each node, its CPU threads, memory and guest storage next to what the guests on it have been given: their cores, memory and disks. What the host uses right now is shown too, but it says little about what the node can still take. A guest that sleeps today has still been sold. The "PNLCS" column counts the guests PNLCS set up; templates are left out. Only guests the API token can see are counted, so a token limited to a pool shows that pool. Below the nodes, the page counts the server's services in PNLCS by status and lists the addresses of the server's IPv4 pool in use: each address, whether it is a server's own or an additional one, and its service and client.
Capacity alert (%) on the server (80 unless set) is the line: once an
hour, pnlcs:proxmox-capacity checks every active Proxmox server and, when a
node has given out that share of its memory or disk, raises
server.capacity_high. Add a notification rule for that event to get it by
e-mail or chat. It comes once a day per node and resource, so there is time to
lower the stock of the products sold there or add a server before an order
cannot be set up. Set the line over 100 if you oversell on purpose, or to 0 to
turn it off. CPU is shown but never alerts: a VPS host shares cores between
guests by design.
Where to look when something went wrong¶
- The service page shows the reason of the last failed job in red.
- The application log (
storage/logs/laravel-<date>.log) has a line for every create, suspend, unsuspend, terminate, upgrade, password change and link, and for every finished reinstall, restore and rollback, with the outcome and Proxmox's message. - Utilities → Activity Log records who pressed which button in the VPS panel — customer or staff — and whether it worked.
- In Proxmox, the task list (Datacenter → Tasks) has the same tasks with their full logs.