This document explains the lifecycle of a Compute Engine instance, coveringthe various states it can go through from creation to deletion. To learn howto check the state of one or more instances, see the following:

• View a list of instances

• View the details of a single instance

By understanding the lifecycle of an instance, you can do the following moreeffectively:

• Troubleshoot instance issues.

• Manage instance resources.

• Plan instance migrations.

Instance states

A compute instance can transition through different states as part of itslifecycle. When creating an instance, Compute Engine provisionsresources to start it, after which the instance moves into staging and preparesfor first boot. After the instance starts, it's considered running. A runninginstance can be repeatedly stopped and restarted, or suspended and resumed,until its deletion.

The following diagram shows the different states that Compute Engine canset an instance to:

The states shown in the preceding diagram are as follows:

• PENDING: after you create a standaloneFlex-start VMwith a wait time (requestValidForDuration) that is 90 seconds or longer,the VM state changes toPENDING. In this state, Compute Engineattempts to acquire the necessary resources to start your VM until its waittime ends. If Compute Engine acquires the resources within this timeand you have sufficient quota for those resources, then the VM state changestoPROVISIONING. Otherwise, you encounter an error andCompute Engine deletes the Flex-start VM. You canoptionally delete the Flex-start VM before the wait time endsif you no longer need the VM.

• PROVISIONING: after you create, restart, or resume an instance,Compute Engine allocates resources for the instance.

• STAGING: Compute Engine is preparing the instance for firstboot due to one of the following reasons:

  • Compute Engine is still creating and configuring the instance.

  • You, or a scheduled operation, have requested to restart or resume theinstance.

  In this state, the instance isn't running yet.

• RUNNING: Compute Engine is booting up the instance, or theinstance is running. In this state, you can stop, suspend, or delete theinstance. Additionally, Compute Engine can stop or delete theinstance for scheduled operations, or repair the instance if a hardwareerror occurs and the instance is part of a managed instance group (MIG).

• PENDING_STOP: the instance is gracefully shutting down. This shutdownprocess happens only if you've enabledgraceful shutdown andyou've requested to stop or delete the instance, or Compute Engineis doing so automatically for a scheduled stop or deletion. The instancestate changes toSTOPPING when one of the following happens:

  • You manually end the graceful shutdown.

  • The graceful shutdown period times out. If any tasks are still running,then Compute Engine forcefully stops them.

• STOPPING: the instance is shutting down its guest OS, which happens inthe following scenarios:

  • You, or a scheduled operation, have requested to stop or delete theinstance.

  • A hardware error occurred.

  Unless you configure the instance toskip the guest OS shutdown,the shutdown time depends on theinstance type.However, if the instance stops due to a hardware error, the shutdown timemight differ from the expected length. After the guest OS shuts down andbased on the operation that is running, Compute Engine does one ofthe following:

  • Compute Engine completes the stops operation and changes theinstance state toTERMINATED.

  • Compute Engine deletes the instance and all attached resources.

• TERMINATED: Compute Engine has completed the stop operation.The attached resources remain attached unless you detach them. In thisstate, the instance remains stopped until you restart or delete it. If yourequest to restart the instance, but Compute Engine can't allocateyour requested resources, then the restart request fails and the instanceremains in theTERMINATED state. Otherwise, the restart request succeedsand the instance state changes toPROVISIONING.

• REPAIRING: Compute Engine isrepairing the instance.Compute Engine repairs an instance if it encounters an internalerror or the instance's host server is unavailable due to maintenance. Whilean instance is in repair, the following happens:

  • You can't use the instance.

  • Theservice level agreement (SLA) doesn't cover theinstance.

  If Compute Engine successfully repairs the instance, then it returnsthe instance state to its original state before the repair operation began.This state can beSTAGING,RUNNING,SUSPENDING, orSTOPPING. If yourinstance is configured to automatically restart (automaticRestart) afterthe repair operation completes, then you can optionally stop the instanceduring the repair process. This action prevents the instance fromautomatically restarting after the repair completes, leaving the instance intheTERMINATED state.

• SUSPENDING: Compute Engine has started the suspend operationof the instance after you've requested to suspend it. In this state, you canonly wait for the suspend operation to complete.

• SUSPENDED: Compute Engine has completed the suspend operation.In this state, you can resume or delete the instance. If you request toresume the instance, but Compute Engine can't allocate yourrequested resources, then the resume request fails and the instance remainsin theSUSPENDED state. Otherwise, the resume request succeeds and theinstance state changes toPROVISIONING. The instance can remain in theSUSPENDED state for up to 60 days. After that time, Compute Enginechanges the instance state toTERMINATED.

Hardware failure

Rarely, a compute instance might fail due to an unexpected outage, hardwareerror, or another system issue. Google recommends mitigating hardware failuresby using persistent storage volumes, routinely backing up your data, anddesigning your system so that a single instance failure isn't catastrophic. Formore information, see how todesign robust systems.

If an instance fails, then Compute Engine automatically restarts theinstance using the same boot disk, metadata, and instance settings. To modifythe automatic restart behavior of an instance, seeSet the host maintenance policy for a compute instance.

Pricing

You're charged for a compute instance as follows:

• For CPU usage, you're charged when the instance is in the following states:

  • RUNNING

  • PENDING_STOP

• For memory usage, you're charged when the instance is in the followingstates:

  • RUNNING

  • PENDING_STOP

  • SUSPENDING

  • SUSPENDED

• For attached resources like disks or external IP addresses, you're chargeduntil the resources exist, regardless of the instance state.

For more information, seeVM instance pricing.

What's next

• View the state of one or more compute instances by doing the following:

  • View a list of instances

  • View the details of a single instance

• Suspend, stop, or reset instances

• Live migration during maintenance events

• Delete an instance