Users Guide

Best practices
Specific guidelines and recommendations for using the SRA and replication software in conjunction with the VMware SRM
disaster recovery solution include the following:
● Prepare a plan in advance for how you will re-establish replication schedules in the event of a site failover. After performing
a reverse-replication operation, you must set up replication schedules in order to ensure periodic replication of data from the
new source volumes back to the original source site. Alternatively, you can initiate replication manually if appropriate.
● Try to group virtual machines with related backup requirements or schedules on the same datastore volume, since replication
occurs on a per-volume basis. For example, if some virtual machines do not need to be replicated to a remote site, or need to
be replicated less frequently, do not store them on the same datastore volume as virtual machines which must be replicated
frequently, to avoid replicating data unnecessarily.
● The SRA only supports replication between identical hardware models. For example, replication between an iSCSI-only
system and a FC/iSCSI hybrid system is not supported.
● Avoid mapping replication volumes to LUN 0 to avoid issues with dynamically mapping and unmapping LUNs, due to special
management functionality assigned to LUN 0. You can map volumes to LUN 0 if those volumes are not expected to be
mapped and unmapped automatically the way replication volumes are, such as local datastores that are not replicated.
● Replication volumes should be mapped with the same LUN number on all hosts.
● Do not use the same LUN number for different volumes that are mapped to different hosts.
● Failover operations will cause read-write host mappings for replication volumes to be converted to read-only, and restoring
replication will convert all read-only mappings for the same volume to read-write. Be careful not to create read-only
mappings of replication volumes such as for data mining purposes. If a read-only mapping of a replication volume is required,
consider creating a non-replicated hardware or software snapshot of the volume.
● The SRA might create host entries on the storage system to keep track of remote IP or FC addresses. Do not delete host
entries whose name starts with “SRA.” However, you may rename them to be more descriptive.
● Replication set basenames for replicated volumes (assigned when creating the replication set) should be no more than 23
bytes long to allow for suffixes to be appended when creating replication snapshots. 23 bytes allows for up to 23 ASCII
characters, but non-ASCII UTF-8 characters require more than one byte each.
● Do not change the name of replication snapshots or change replication-set basenames except to conform to these best
practices. The SRA depends on consistency between the replication-set basenames and the snapshot names.
4
Best practices 13