Search

22.5. Basic Example: Performing a Recovery

download PDF
An administrator, John Smith, has to create a disaster recovery plan for his directory deployment. Example Corp. has three physical offices, in San Francisco, Dallas, and Arlington. Each site has 10 servers which replicate to each other locally, and then one server at each site replicates to another server at the other two sites.
Each site has business-critical customer data stored in its directory, as well as human resources data. Several external applications require access to the data to perform operations like billing.
John Smith's first step is to perform a site survey. He is looking for three things: what his directory usage is (clients that access it and traffic loads across the sites), what his current assets are, and what assets he may need to acquire. This is much like the initial site survey he performed when deploying Red Hat Directory Server.
His next step is identifying potential disaster scenarios. Two of the three sites are highly vulnerable to natural disasters (San Francisco and Dallas). All three sites could face normal interruptions, like outages for power or Internet access. Additionally, since each site suppliers its own local data, each site is vulnerable to losing a server instance or machine.
John Smith then breaks his disaster recovery plan into three parts:
  • Plan A covers losing a single instance of Directory Server
  • Plan B covers some kind of data corruption or attack
  • Plan C covers losing an entire office
For plans A and B, John Smith decides to use a hot recovery to immediately switch functionality from a single instance to the backup. Each server is backed up daily, using a cron job, and then the archive is copied over and restored on a virtual machine. The virtual machine is kept on a different subnet, but can be switched over immediately if its peer ever does offline. John Smith uses simple SNMP traps to track each Directory Server instance's availability.
Plan C is more extensive. Along with replication between sites and the local backups, he decides to mail a physical copy of each site's backup, for every local instance, once a week to the other two colocation facilities. He also keep a spare server with adequate Internet access and software licenses to restore an entire site, using virtual machines, one of the other different colocation facilities. He designates the Arlington site as the primary recovery location because that is where most of the IT staff is located, then San Francisco and last Dallas, based on the distribution of personnel. For every event, the IT administrator at all three sites will be notified, and the manager assumes the responsibilities of setting up the virtual machines, restoring the Directory Server instances from the physical backups, and rerouting client traffic.
John Smith schedules to review and update the plan quarterly to account for any new hardware or application changes. Once a year, all three sites have to run through the procedure of recovering and deploying the other two sites, according to the procedures in Disaster Plan C.
Red Hat logoGithubRedditYoutubeTwitter

Learn

Try, buy, & sell

Communities

About Red Hat Documentation

We help Red Hat users innovate and achieve their goals with our products and services with content they can trust.

Making open source more inclusive

Red Hat is committed to replacing problematic language in our code, documentation, and web properties. For more details, see the Red Hat Blog.

About Red Hat

We deliver hardened solutions that make it easier for enterprises to work across platforms and environments, from the core datacenter to the network edge.

© 2024 Red Hat, Inc.