Go to main content

Using Unified Archives for System Recovery and Cloning in Oracle® Solaris 11.3

Exit Print View

Updated: October 2017

Best Practices Using Unified Archives

This section covers the following topics:

Improving Data Consistency With Multiple Pools

Archive components are consistent with one another, meaning that when an archive is deployed, all data present post-deployment should be from the same point in time from the origin system's perspective. Consistency is guaranteed on a per-pool basis. However, if an Oracle Solaris instance spans multiple pools, as when a root pool and a separate application or data pool are used, you should quiesce any active workloads prior to archive creation to ensure consistency between the pools once they are deployed.

Data Preparation

By default, all datasets other than swap and dump volumes are included when a Unified Archive is created. Datasets that are not needed for redeployment may be excluded using the exclude-dataset option (–D) to the archiveadm create command. If you exclude one or more datasets, the archive will be smaller and will be created faster. Because directory and file exclusion lists are not provided, remove any temporary data or data that is otherwise not needed for redeployment prior to archive creation. When possible, place optional temporary data directories or any other data not needed for redeployment in their own datasets so that you can use the exclude-dataset option.

Naming Unified Archives

When creating Unified Archives, use a file name that provides information about the archive. For example, noscratch-recovery.uar could tell you that the file is a recovery archive of a system that doesn't include the scratch dataset. Note that the name of the system that the archive was created from, the deployable systems in the archive, and the date the archive was created are stored in the metadata of the archive, which you can display using the archiveadm info command. You can see additional information by adding the –v option to this command. For examples showing the command output, see Viewing Unified Archive Information.

SVR4 Packages and Clone Archives

The IPS packaging system makes certain tasks available during archive creation that weren't possible with SVR4 packages. For example, the creation process for recovery archives cleans up /dev and prepares the mappings of physical device names to instance numbers in /etc/path_to_inst. The creation process for clone archives performs the same operations and also cleans up other data that is specific to an Oracle Solaris instance, such as removing system configuration information. Unified archives use tagging attributes defined in IPS manifests to perform these tasks. These attributes also enable users to tag their own IPS packages for use when creating an archive.

To take advantage of IPS functionality during archive creation, convert SVR4 packages to IPS packages and install the packages onto a ZFS dataset. The archiveadm command only works on data that is managed by ZFS. For more information, see Converting SVR4 Packages To IPS Packages in Packaging and Delivering Software With the Image Packaging System in Oracle Solaris 11.3.