# Troubleshooting Compute

Symptoms you can hit when launching and running instances, volumes and addresses, with their cause and fix.

## Launch is refused: "Missing required tags: …"

**Cause:** your organization requires those tag keys on new resources.

**Fix:** add the named keys to `tags` (in the console, under **Advanced** → **Tags**) and launch
again.

## Launch is refused: "Subnet … has no free addresses left"

**Cause:** every address in the subnet's range is taken.

**Fix:** launch into another subnet, or into a network with a larger range.

## Launch or resize is refused with 409 because the type is out of stock

**Cause:** the type has no stock in that zone right now.

**Fix:** for a launch, use a zone the message names as having it; otherwise pick a type in stock;
see
[Check stock before you launch](check-stock.md).

## Launch or resize is refused: "Invalid or unavailable server type"

**Cause:** the type is not in the catalog, or is deprecated.

**Fix:** pick a type from the launch form, or from the `dcloud_server_types` data source.

## The instance stays in `pending`

**Cause:** the machine is still being created and booted.

**Fix:** wait; it moves to `running` on its own. `dcloud ec2 instances wait` returns when it does.

## SSH is refused or times out

**Cause:** no key pair was chosen at launch, so no key is installed; or no security group lets port
22 in from your address.

**Fix:** add a key through Connect (the instance page offers **Add a key**), or use the browser
terminal; check the instance's security groups on its **Networking** tab.

## Start is refused: "Cannot start instance in state: running"

**Cause:** the action does not match the instance's state.

**Fix:** `start` needs `stopped` (or `suspended`); `stop`, `shutdown` and `reboot` need `running`.
Wait until a transition (`pending`, `stopping`, `rebooting`) has finished.

## Resize is refused: "Instance must be stopped before resizing"

**Cause:** the instance is running.

**Fix:** stop it, resize, then start it.

## Attaching a volume is refused: "… is in … and … is in …. A volume attaches only to an instance in its own availability zone"

**Cause:** a volume attaches only to an instance in its own zone.

**Fix:** launch the instance in the volume's zone, or snapshot the data into a volume in the
instance's zone.

## Attaching an encrypted volume is refused: "… was launched without the volume helper"

**Cause:** the volume helper was not installed at launch: the instance is older than the helper, or
its user data left no room for the helper within the 32 KiB limit.

**Fix:** install it, launch a new instance, or create the volume with encryption off; see
[Use encrypted volumes](encrypted-volumes.md).

## An associated elastic IP does not answer

**Cause:** an instance launched before 17 November 2026 does not bind the address itself (nor does
one whose user data left no room for the binding). A newer instance binds it within 20 seconds.

**Fix:** run the commands the address shows under **On the machine** on **Elastic IPs**, once, as
root (written for Ubuntu or Debian with netplan).

## Release is refused: "Disassociate the IP from its instance before releasing"

**Cause:** the address is still attached.

**Fix:** disassociate it first, then release it.

## A `terminated` instance is still listed

**Cause:** the record is kept after the machine is deleted.

**Fix:** terminate it once more (**Delete record** in the console) to remove it from the list.

## Deleting a network is refused: "Delete what is still in this VPC first: …"

**Cause:** instances, databases, caches, VPN gateways or a NAT gateway are still in it.

**Fix:** delete what the message names, then delete the network. The default VPC cannot be deleted
at all ("Cannot delete the default VPC").
