Thales recommends using auto-recovery for all HA group configurations (see Configuring HA Auto-Recovery). If you do not enable auto-recovery and a member partition fails, or if the recovery retry count expires before the partition comes back online, you must recover the partition manually using LunaCM. You do not need to pause your application(s) to perform a manual recovery; the HA group handles load-balancing and automatically replicates any new or changed keys to the recovered member.
To perform a manual recovery of a failed HA group member
1.[Optional] Ensure that the failed member is available and visible in LunaCM by addressing the problem that caused the failure. Display the HA group to see the failed members. You are prompted for the Crypto Officer password/challenge secret.
lunacm:> hagroup listgroups
HA Group Label: myHAgroup
HA Group Number: 1154438865287
HA Group Slot ID: 5
Synchronization: enabled
Group Members: 154438865287, 1238700701509
Needs sync: no
Standby Members: <none>
Slot # Member S/N Member Label Status
====== ========== ============ ======
------ 154438865287 par0 alive
------ 1238700701509 ------------ down
2.If you are using a multifactor quorum-authenticated partition with auto-activation disabled, or if the partition was down for longer than two hours, log in to the partition as Crypto Officer and present the black CO PED key.
lunacm:> slot set -slot <slotnum>
lunacm:> role login -name co
3.Execute the manual recovery command, specifying the HA group label.
lunacm:>hagroup recover
If you have an application running on the HA group, the failed members will be recovered the next time an operation is scheduled. Load-balancing and key replication are automatic.
4.If you do not currently have an application running, you can manually synchronize the contents of the HA group.
CAUTION! If manual synchronization is used while an application is running, two processes will attempt to clone the same missing object simultaneously. While one process will succeed in the cloning operation, the other will fail and generate error messages in the logs. These errors are harmless, but they can be a cause of confusion and alarm.
CAUTION! Object deletion can be disrupted when another process is running a manual synchronization. As a result, the object may not be completely removed from the HA group. If the HA group undergoes another manual synchronization later, the object that got left behind may be replicated back to all HA members, reintroducing an object that was intended to be removed.
lunacm:> hagroup synchronize -group <label>
