Fix: a provider rejects an action or reports it busy

What you're seeing

A customer quotes The operation could not be completed. The error has been recorded under reference …, or is told The server is still finishing another operation. Please try again in a minute or two. The Dashboard may show an alert for a failed action or setup.

How ImpulseVPS reports a refusal

When a provider refuses a request, that request is not carried out (a rate-limited create is the exception; see below).

• Your customer gets a plain message: what to do when the fix is theirs, otherwise the generic message with a reference, never the provider's error. On a product that hides its provider, provider names are replaced with "your cloud server".

• You get the provider's error message, with its HTTP status when the provider answered, on the service's page under Customers (Recent actions, or Last error for a failed setup), and in Diagnostics → Logs and Diagnostics → Activity. The customer's reference is the request ID in that action's Activity entries, which you can filter by service number.

The token is missing access

Test connection and the 15-minute provider check only read from the provider. A token that can't write still passes both, so the first sign is a failed order or action:

• Hetzner Cloud and DigitalOcean tokens need read and write access, not read only.

• Linode tokens need Linodes, Volumes, StackScripts, IPs and Firewalls read/write. Without Firewalls, customers' firewall changes fail.

• Vultr sub-user keys have only that user's permissions. The account's Personal Access Token has full access.

A key the provider rejects outright raises a critical alert, The API key for … was refused: the key is wrong, revoked or lacks read access, or Vultr's API access control doesn't allow your WHMCS server's IP. Every order and action on that account fails until it's fixed.

1. Create a token with that access. Linode can't add a scope to an existing token, so create a new one.

2. Paste it into the same account under ImpulseCore → Operations → VPS Providers. Servers stay with the account they were created on, so don't add a new account.

3. Click Test connection on the account in Addons → ImpulseVPS → Settings → Providers & Regions. To prove the token can create servers, run the Synthetic end-to-end test under Diagnostics → Health Checks (a real, billable server for a few minutes). It does not test firewall, reverse DNS or volume access, so for those, have the customer retry the action that failed.

4. Have the customer retry. If setup failed, run Module Create again on the service.

For the access each provider's token needs, see Adding a provider account to ImpulseVPS.

The provider says the server is busy

All four providers refuse a change while the server is finishing another one, and some jobs take minutes: Linode won't power off a server during a reinstall, and a DigitalOcean power change can wait minutes in its queue.

• Power on, off and reboot, firewall and reverse DNS changes answer at once. A busy refusal asks the customer to retry. Three busy refusals on one service within 30 minutes raise a warning alert, because the server may be locked at the provider.

• An accepted power-on or power-off stays in progress, never sent twice, until the provider reports it done or the power deadline (15 minutes by default) fails it.

• Reinstalls, snapshot creates and restores, and volume creates, attaches and detaches stay queued and are tried again each time the scheduled tasks run (every 5 minutes with the standard WHMCS cron). Until then, a new power, reinstall, snapshot or volume action on that server is refused with Another operation is already in progress on this server — please wait for it to finish. If the provider is still busy at the deadline (by default 30 minutes for a reinstall, 60 for a snapshot), the action fails, the server is unlocked, and you get a critical alert.

• Suspend re-sends the power-off while the provider refuses it and waits up to three minutes for the server to be confirmed off. If the server is still busy, WHMCS reports the suspension is incomplete, nothing is destroyed, and you get a critical alert. Suspend again once the job finishes.

Usually, waiting fixes it. If an alert says a server may be locked, check it in the provider account. Deadlines are under Settings → Alerts & Notifications.

Rate limits

ImpulseVPS has no special handling for rate limits: a server action refused with HTTP 429 fails like any other provider error.

A rate-limited create is different, because the answer doesn't say whether a server was made. The service is held as an unknown create, and you get a critical alert:

1. In the provider account, look for a server labelled impulse-vps-N or tagged svc-N, where N is the service number. Destroy it if it's there.

2. Click Resolve unknown create on the service's page under Customers.

3. Run Module Create again.

Linode reverse DNS needs forward DNS first

Linode sets reverse DNS only for a name that already points to the address. The customer is told Reverse DNS can be set to <name> only once that name points to this server, with the A or AAAA record to add, and you get no alert. If the default reverse DNS that setup tries to set is refused, the server is still delivered, the refusal is logged as rdns.default_failed, and the customer can set it later on the Network tab.

Vultr volumes have a minimum size

On Vultr, NVMe volumes start at 10 GB and standard-tier volumes start at 40 GB. Dallas, Miami and San Jose are among the locations that offer only the standard tier. When ImpulseVPS can read the location's storage tiers, it checks first, so a customer asking for less is told Storage volumes in this location start at 40 GB. If it can't read them, the create fails afterwards and you get a failed-action alert saying the customer needs to choose 40 GB or more. Additional Storage is sold in 10 GB blocks. On server plans that sell those locations, set the storage limit to at least 40 GB and put the minimum in the product description. For more on volumes, see How snapshots, backups and extra storage work.

How to confirm it worked

The customer's retry succeeds. A failed-action alert closes when that action next succeeds on the service, a busy-server alert when any change to the server succeeds, and a refused-key alert about 45 minutes after the key works. A failed-action alert that stays open has Mark resolved: use it after checking the provider account.

Still stuck?

Open a ticket from the portal at www.impulsesoftware.com with the provider, the service number and the provider's error from Recent actions, and attach the support snapshot from Diagnostics → Help (it masks secrets).

Related

• Adding a provider account to ImpulseVPS

• How snapshots, backups and extra storage work

• How to monitor your ImpulseVPS fleet