Introduction

While OpenStack is a resilient platform in High Availability (HA) deployments, disasters that cause a hardware node or an entire cloud to fail may still occur. Due to the potential for unexpected events to cause your business or operations to fail, we suggest planning for disaster events. In this guide, we emphasize the importance of planning for disasters and outline strategies to consider when creating your disaster recovery plan.

Disaster Recovery Strategies

Disaster recovery varies greatly depending on the needs of your organization. Your disaster recovery plan should involve analyzing your company’s most valuable data. This audit should include an inventory of documents, databases, and systems that are deeply involved in the revenue-generating aspects of your business. As an infrastructure provider, we also have the responsibility to maintain the integrity of our facilities to the best standards and practices to safeguard your data. In addition to our responsibilities, you should take the additional steps necessary to protect the integrity of your data against potential impacts on your private cloud.

Recovery Objectives

The objectives behind a disaster recovery plan fall into two categories:

  • Recovery Point Objectives – This is a specific point in time that data must be backed up to resume business.
  • Recovery Time Objectives – Length of time that a system can be offline before the business is negatively impacted.

Consider the above objectives when determining the needs of your disaster recovery strategies. In recovery point objectives, determining the frequency of your backups is critical to your disaster strategy. Recovery time objectives are based on the amount of lost revenue per unit of lost time. This means that every hour or minute that certain systems are offline can greatly impact a business.

Off-site Backups

To have an effective disaster recovery plan, we stress the use of an off-site backup solution. Off-site backups are a critical component to preventing catastrophic loss of data and limiting the blast radius of an outage. Although some services in OpenMetal are highly available, high availability does not protect against unforeseen natural disasters. In the event of a hurricane, flood, or other unforeseen circumstance, it could be possible that you must recover data from a different geographical location. It is highly recommended that off-site backups are used for the most mission-critical data.

RBD Mirroring with Ceph

Private Clouds by default use Ceph as a shared storage backend. As part of your disaster recovery plan, consider using Ceph’s RBD mirroring feature to mirror important data to another Ceph cluster, in a geographically different data center location. Note that we currently only provide one data center in which to host your Private Cloud.

Handling a Hardware Failure

Great care is taken to monitor and maintain our infrastructure. However, in the event of hardware failure, it might be necessary to diagnose these issues. If you believe you are experiencing hardware failure, contact our support. Two possible hardware-related issues are failure of a hardware node and failure of an entire cloud. The following sections outline how to diagnose hardware failure.

Determine Hardware Node Failure

To quickly get the status of your cloud’s control plane nodes, check the running system services in Horizon. To do so, log in as an administrator to Horizon and navigate to Admin -> System -> System Information. On this page are tabs separating all OpenStack services, including the APIs, Compute, Block Storage, and Network Agents. A service’s state is indicated by the State column. If a service is up, the value is reflected as “Up”. If a service is down, the value is reported as “Down”. Click through each tab to see each service’s status. A good indicator that a control plane node is down is if you see “Down” under the State column for all services of a particular node.

Figure 1: OpenStack System Information

Cluster Failure

If your entire Private Cloud has failed, the signs will be readily apparent. If you are using monitoring software such as Datadog or Nagios, you will be alerted to your nodes being offline. In addition to being alerted, you will be unable to access your assets.

Note: In some instances, it is also possible to see a failure of your nodes within OpenMetal central. To view the status of these nodes, go to the assets page of OpenMetal central, and under the hardware section, there is a green indicator icon reflecting the current status of your nodes. If the icon is yellow or red, then the issue is likely hardware-related.

Figure 2: Assets Page of OpenMetal Central

Cloud Monitoring with Datadog

Datadog is an optional cloud monitoring service we provide for OpenMetal Private Clouds. Should you want to add this feature to your cloud, contact your Account Manager or submit a support ticket in OpenMetal Central.

Contact Support

If you are experiencing hardware failure or any other issues with your Open Metal Private Cloud, Contact Support.

Additional Reading

For more regarding OpenStack Disaster Recovery, see: