1 Pictures Second Understanding Nova 16 operations-5 minutes a day to play OpenStack (44)

we discussed some of the operations of Instance in the previous some The operation function is similar, also has the respective application scene, now is the time to summarize systematically.

As shown, we divide the management of Instance into two categories: normal operation and fault handling.

General Operations

In normal operation, Launch, Start, Reboot, Shut Off and Terminate are all well understood. Here are a few things to look back on:

Resize
Adjust the resources assigned to instance by applying different flavor.

Lock/unlock
Can prevent the misoperation of the instance.

Pause/suspend/resume
Pauses the current instance and resumes at a later time. The difference between pause and Suspend is that pause saves the running state of instance in the memory of the compute node, and Suspend is saved on disk. The advantage of Pause is that the resume is faster than Suspend, and the disadvantage is that if the compute node restarts and the memory data is lost, the resume is not available, and Suspend does not have the problem.

Snapshot
Backup instance to Glance. The resulting image can be used for failback, or to deploy a new instance for the template.

Fault Handling

There are two scenarios of fault handling: planned and unplanned.

Plan refers to the maintenance work scheduled in advance of the time window, such as Server regular microcode upgrade, add replacement hardware and so on. Unplanned unexpected failures, such as forced shutdown caused the OS system file corruption, server power down, hardware failure.

in-plan fault handling

For planned failure handling, you can migrate instance to other compute nodes in the Maintenance window. The following actions are involved:

Migrate
Migrate instance to other compute nodes. Before migration, instance will be Shut Off, supporting both shared and unshared storage.

Live Migrate
Unlike Migrate, live Migrate is able to migrate instance online without downtime, ensuring continuity of business. Shared storage and non-shared storage (Block migration) are also supported

Shelve/unshelve Shelve saves instance to Glance and can then be redeployed through Unshelve. After the Shelve operation succeeds, the instance is removed from the original compute node. Unshelve will re-select the node deployment, which may not be the original node.

Unplanned fault Handling

Unplanned failures fall into two categories of impact: Instance failures and compute node failures

Instance Fault

Instance fault is limited to one Instance operating system level, the system does not start normally. You can repair instance with the following actions:

Rescue/unrescue
Starts with the specified boot disk, enters Rescue mode, and repairs the damaged system disk. After a successful repair, instance is started normally by Unrescue.

Rebuild
If Rescue cannot be repaired, it can only be recovered from an existing backup through Rebuild. Instance backups are created through snapshot, so backup policies are required to be backed up regularly.

Compute node Failure

The Instance fault is confined to a specific Instance, and the compute node itself is working properly. If the compute node fails, OpenStack cannot communicate with the node's nova-compute, and all instance running on it will be affected. At this point, the Instance can only be rebuilt on other normal nodes through the evacuate operation.

Evacuate
Rebuilds Instance on other compute nodes using the Instance image file on the shared storage. So planning for shared storage ahead of time is key.

Section

Here, we've learned about the OpenStack Nova architecture, discussed key components such as Nova Api,scheduler,compute, and analyzed the Nova operations in detail, and finally summed up the purpose and usage scenarios for these operations using a single graph.

Nova is the most important project in OpenStack, at the center of OpenStack. Other Keystone,glance,cinder and Neutron projects are for Nova's service, so be sure to understand them well.

In the next section we will learn about the OpenStack Block storage service-Cinder.

1 Pictures Second Understanding Nova 16 operations-5 minutes a day to play OpenStack (44)