Atlas Docs

VM stuck in Starting

Properties1
descriptionAn instance never progresses past pending / Starting.

2 min read

Symptom

You launched an instance — via Console, cmk, Terraform, or the API — and it never progresses past Starting. The instance state stays Starting (or Pending / Allocated) indefinitely. In the Console, the spinner doesn’t resolve. From the CLI:

$ cmk list virtualmachines name=hello-atlas filter=state
state = Starting

Investigation

Work through these in order — the answer is almost always one of these four.

1. Zone capacity

The most common cause is the zone running out of capacity for the Compute Offering you picked. Check whether other instances of the same size are scheduling normally — try launching a second one in the same zone. If it also stalls, capacity is the issue.

2. Your account quota

Atlas accounts have per-resource quotas (max VMs, max vCPUs, max RAM). The instance may be silently waiting for quota approval. Check the Console Account → Resource limits view, or cmk list resourcelimits. Hitting a quota usually produces an error rather than a stall, but quota errors are sometimes deferred.

3. Template availability

If the Template you selected is being re-uploaded or re-registered, instances can stall until it’s ready. Confirm the template state:

cmk list templates id=<TEMPLATE_ID> filter=isready

isready = true means it’s good. false means wait or pick another.

4. Network attachment failure

If the Guest Network or VPC tier you attached has a problem (e.g. the IP range is exhausted, or the network is itself stuck), the instance can stall while waiting for an IP. Try detaching the network attachment and re-launching with a different network.

Solution

Most often:

  • If the issue is zone capacity or template re-upload: wait 5-10 minutes and check state again. These usually resolve themselves.
  • If the issue is quota: request a quota raise via mailto:help@runatlas.is or pick a smaller Compute Offering.
  • If the issue is network: stop the instance, detach the bad network, attach a different network, start.

If the instance is still stuck after 15 minutes and none of the above apply, destroy and re-create it. CloudStack does not always report scheduling failures cleanly; destroy+re-create is the canonical reset.

If it stalls a second time on a fresh launch, email mailto:help@runatlas.is with the instance UUID and the timestamp — that’s a platform-side issue.

Backlinks