Administrator Guide

NOTE: If NFS exports are used, the NAS volume names of the source and target should be the same, as the export
path name includes the NAS volume name. This is not relevant for SMB shares.
…………………………
Source volume An (Cluster A) to target volume Bn (Cluster B)
3. Ensure that at least one successful replication has occurred for all the source volumes in Cluster A.
If the replication fails, x the problems encountered and restart the replication process.
4. Record all Cluster A settings for future use. Replication restore is not a complete BMR (bare metal restore). Settings such as
network conguration (client, SAN, and internal) cannot be backed up and restored using the replication method. Note all
Cluster A settings (for use when restoring Cluster A) including network conguration, cluster wide settings such as cluster
name, alert settings, and so on for future use. If the system restore operation fails to restore these settings, you can manually
restore the Cluster A settings back to their original values.
Phase 2 — Cluster A fails and clients request failover to target Cluster B
If Cluster A stops responding because of an unexpected failure, fail over to Cluster B.
1. From Cluster B, promote the target volumes in Cluster B. This transforms the original target volumes (B1, B2, .. Bn) to
standalone NAS volumes and makes them writable.
2. Delete the replication policies for the original source volumes (A1, A2, .., An).
3. Apply the source volume conguration from the original source volumes in Cluster A to the target volumes in Cluster B.
4. Restore the users and groups conguration from Cluster A. This restores the Cluster B users and groups to Cluster A settings.
5. Ensure that Cluster B is used to temporarily serve client requests during the failover time.
a. Choose one of the following options:
• IP address-based failovers: Change the IP addresses for Cluster B to match the IP addresses used by Cluster A. Existing
client connections might break and might need to be re-established.
• DNS-based failovers: Point the DNS names from your DNS server to Cluster B instead of Cluster A.
Ensure that the DNS server on Cluster B is the same as the DNS server or in the same DNS farm as the DNS server of
Cluster A. Existing client connections might break and might need to be re-established. You must unmount and re-mount
the NFS exports on the clients.
b. (Single NAS volume failovers) Manually update the DNS entry for the NAS volume that was failed over. This redirects clients
that are accessing this volume from Cluster A to Cluster B, while other clients keep accessing other volumes using the same
DNS name. Client systems might need to refresh their DNS cache.
c. (Single NAS volume failovers) To force SMB and NFS clients to Cluster B, you must delete the SMB shares and NFS
exports on Cluster A. This forces the SMB and NFS clients to reconnect, at such time they are connected to Cluster B.
After restoring the source volume’s conguration on Cluster B, all of the SMB shares and NFS exports will be present on the
target volume (on Cluster B), so no SMB share/NFS export conguration information is lost.
The failed over volume can now be accessed using the exact same DNS name and SMB share/NFS export name as it was
when hosted on Cluster A, except now it is hosted on Cluster B.
d. Join Cluster B to the AD server or LDAP/NIS.
Ensure that the AD server and LDAP server are in the same AD/LDAP farm or same server.
Phase 3 — Restore Cluster A and fail back from Cluster B to Cluster A
After you have xed the reason that caused Cluster A to fail, fail back over to Cluster A.
1. Fix the reason that caused Cluster A to fail and if required reinstall FluidFS.
2. Rebuild the FluidFS cluster:
• IP address-based failovers: Use the settings for Cluster A that you recorded earlier, but change the IP addresses for Cluster
A to match the IP addresses originally used by Cluster B.
• DNS-based failovers: Use the settings for Cluster A that you recorded earlier.
3. From Cluster B, set up a replication partnership between Cluster B and Cluster A.
4. Congure replication for all the promoted recovery volumes in Cluster B, and specify that they replicate back to the original
source volumes in Cluster A.
The replication policy must be a one-to-one match on a volume basis, for example:
458
FluidFS Data Protection