Launching an instance corresponds to both creating it and then starting it. Once the instance is launched, it enters the pending state until it is created, started and ready to use. The state of the instance then changes to running.
Launching an instance allocates the corresponding physical resources to it. You define the hardware of its host computer thanks to an instance type, and the OS, configuration and possibly software applications installed on it thanks to an OUTSCALE machine image (OMI). For more information, see About OMIs.
The instance receives a unique instance ID, and a private IP address and an associated private DNS name that can only be contacted within the Cloud. If you launch an instance in the public Cloud, it also receives a public IP address and an associated public DNS name. This public IP address is temporary and changes every time you stop and start the instance. For more information, see About Instances > General Information About Instances.
- To fix a public IP address to an instance of the public Cloud or to add a fixed public IP address to an instance in a Virtual Private Cloud (VPC), you can attach an External IP (EIP) to it, and then use the osc.fcu.eip.auto-attach OUTSCALE tag to fix it through the stop and start process. For more information, see About EIPs.
- To launch an instance in the public Cloud without a public IP address and a public DNS name, use the private-only OUTSCALE tag.
For more information, see Configuring an Instance with User Data and OUTSCALE Tags.
You can also attach flexible network interfaces to an instance at launch or after its creation to partition its network. For more information, see Flexible Network Interfaces (FNIs).
Stop and Start
You can stop a running instance at any time, and then start it again. Stopping an instance using the API corresponds to shutting it down using the operating system (OS) command.
You can also force an instance to stop. When doing so, the instance stops without properly exiting running applications.
Forcing an instance to stop may damage its system.
When you stop an instance, it enters the stopping state, and then changes to stopped. When you start a stopped instance, it enters the pending state, and then changes to running.
When stopping and starting an instance, it keeps:
- Its instance ID
- Its private DNS and IP address
- The EIP attached to it and fixed using the osc.fcu.eip.auto-attach OUTSCALE tag (if so)
If the instance is not tagged with the osc.fcu.eip.auto-attach OUTSCALE tag, the EIP is detached from the instance when stopping it.
- Volumes attached to it (if so)
However, the public IP address and public DNS name assigned to an instance of the public Cloud at launch change through the stop and start process. Data stored in the memory are also erased when stopping the instance.
When an instance is stopped, you can modify its attributes like its instance type (amount of vCores and memory). For more information, see Modifying an Instance Attribute.
It is also recommended to stop the instance if you want to treat a volume attached to it. To do so, you need to detach the volume from the instance once it is stopped and to attach it to another instance to treat it. When reattaching the volume to the original stopped instance, beware of using the same device name as the one specified in its block device mapping before starting the instance again. For more information, see Attaching a Volume to an Instance, Detaching a Volume from an Instance and Defining Block Device Mappings. If the instance you want to stop is registered with a load balancer, it is recommended to deregister it before stopping it, and to start it before registering it again if needed. For more information, see Load Balancing Unit (LBU).
If regular stop does not work on your instance, you can also force it to stop. Forcing the instance to stop corresponds to unplugging a computer, which means that the system may not stop properly.
Forcing an instance to stop may damage its system and lose data. Ensure that you no longer need it or that you have backed them up.
You can open the console to check if there is any issue or problem going on. To do so, click Console Output in the Instances page on Cockpit.
The following list presents the general causes why an instance cannot stop properly:
- A process is running, which prevents to stop the instance.
The most frequent reason is that a process is using a filesystem, which means that an operation on the volume is not finished yet. This prevents the filesystem to be unmounted, which is required to stop the instance. Therefore, you need to ensure that no process is using a filesystem (for example, NFS or CRFS) . If there is any, do one of the following three options before trying again to stop the instance:
- Wait until the process ends.
- Stop the process.
- Unmount the volume on which the process is running.
- An update is in progress (for example, a Windows update).
Do not force an instance to stop during an update, as it may damage your instance or prevent it from starting again. Some updates may take a lot of time (up to several hours) on small instance types.
- There are issues with your ACPI calls, required by instances to stop properly. The two main issues are:
- The instance has crashed. In this case ACPI calls are ignored.
- The pci-hotplug and acpiphp modules are not installed on your instance, so ACPI calls are not supported. This may happen if you used your own custom OMI to launch your instance.
You can check the /etc/modules directory to check whether they are installed or not:
If the modules are not installed, you can create your own custom OMI using an official OMI, and then launch a new instance.
If the instance does not stop when using a Force Stop, contact our Support team. Beware that the support will have no other choice but shutting the instance down which may corrupt your data. Ensure you have backed them up.
You can reboot a running instance at any time if needed using the API, which corresponds to rebooting the OS. When rebooting, the instance restarts without going through the stop and start process.
The instance is still running and keeps all its allocated resources. Besides, data stored in its memory remain available after rebooting an instance.
You can terminate an instance that you no longer need. Terminated instances cannot be recovered. The instance enters in the
shutting-down state, and then changes to
terminated once the termination is completed. The instance remains visible in the
terminated state for 1 hour, without any possibility to recover it.
When terminating an instance, its corresponding physical resources are released and data stored in its memory are erased. If an EIP is attached to the instance, it is released but still allocated to your account.
The behavior of BSU volumes when terminating the instance they are attached to depends on the block device mapping. By default, the root device of the instance is deleted while other volumes attached to it are detached. For more information about how to set this behavior, see Defining Block Device Mappings.
You can also configure two types of termination protection:
DisableApiTermination: This attribute enables you to prevent instance termination (by default, enables termination).
InstanceInitiatedShutDownBehavior: This attribute enables you to define the instance behavior when you stop or terminate it. By default or it set to
stop, the instance stops. If set to
restart, the instance stops then automatically restarts. If set to
terminate, the instance stops and is terminated. You can, for example, automatically terminate an instance at the end of an application by setting this attribute to
terminateand asking the instance to stop when the running application is closed.
These two termination protection attributes can be defined when launching the instance and modified afterward. For more information, see Modifying an Instance Attribute.