SQL Always On Subclient Pinned to a Paused Replica
A SQL Server Always On Availability Group has three replicas: SQLA (primary), SQLB (synchronous secondary), and SQLC (asynchronous secondary in the DR site). Commvault SQL iDA protects 40 databases. The subclient is configured to back up from a secondary replica, pinned to SQLB, with a full weekly and transaction log every 15 minutes. During patching, SQLB is paused for 6 hours. Log backups begin failing with 'database is not accessible' and 'preferred replica is not available.' The DBA fails over to SQLC and runs a native full backup on SQLC. A new database, Finance_New, is added to the AG but does not appear in Commvault. The RPO is 15 minutes; the RTO is 2 hours. The team wants to point the subclient at SQLA and stop taking log backups until the next full.
1. Diagnose the log backup failures. What is the role of the preferred replica, and why is pointing the subclient at SQLA and stopping log backups a poor plan?
Read the worked answer
The subclient is pinned to SQLB as the preferred secondary replica. When SQLB is paused, Commvault cannot connect to that replica for the log backup, so jobs fail even though another replica is available. A failover to SQLC does not automatically change a hard-coded preferred replica in the subclient. Pointing the subclient at SQLA may restore backups, but it shifts backup load to the primary and may conflict with the AG backup preference. Stopping log backups until the next full directly violates the 15-minute RPO and leaves a gap that cannot be recovered point-in-time. Also verify what the DBA actually ran: a native full backup does not by itself break the log chain, but a native log backup can create a recovery gap for Commvault's chain. Check the database recovery model and the log chain before assuming the native full is the cause. The better fix is to update the preferred replica configuration or allow backup from any available secondary, then confirm that log backups resume and the chain is continuous.
2. How should Finance_New be protected without manually editing the subclient every time? What auto-discovery and AG-level settings are relevant?
Read the worked answer
Protect at the availability-group or instance level with auto-discovery enabled, rather than maintaining a hand-edited list of database names. Confirm that the Commvault service account has the required permissions on the AG listener and on the new database, and that Finance_New is online and healthy in the AG. Run a discovery or refresh operation and check whether the subclient content is set to include all databases in the AG or only selected databases. If the subclient uses explicit database entries, a new database will not appear until it is added; that is a design choice, not an automatic failure. Use AG-aware discovery and test with a newly added database before relying on it. Also confirm that the preferred replica policy applies to the new database and that log backups are running. A restore test of Finance_New should include a full restore and a point-in-time log restore on a non-production SQL instance.
3. Design the change and restore test that would prevent recurrence and prove the 15-minute RPO can be met.
Read the worked answer
First, reconfigure the SQL subclient so it does not depend on a single paused replica: either update the preferred replica list in a controlled way or allow backup from any eligible secondary. Keep log backups running; do not wait for a new full unless the log chain is actually broken. Test a planned failover from SQLA to SQLC and confirm that log backups continue within the 15-minute interval. Confirm that the AG backup preference and Commvault settings do not conflict. For restore validation, restore the latest full and all available logs to a separate SQL instance, stop at a chosen point in time before and after the failover, and verify application-consistent recovery. Record the actual RPO achieved during the failover, not just the configured schedule. Add monitoring for failed log backups, preferred replica availability, and AG role changes. Document the procedure for pausing a replica so backups are moved before maintenance begins.