MinervaDB Consultative Support: 12 Engines, One Proven 24×7 Team

SQL · PostgreSQL: partial covering index and plan verification

-- PostgreSQL 16+: partial covering index on the hot window, built without blocking writes.
-- Partial-index predicates must be immutable, so the boundary is a literal that a scheduled job rolls forward.
CREATE INDEX CONCURRENTLY idx_orders_recent_status
    ON orders (status, created_at DESC)
    INCLUDE (customer_id, total_amount)
    WHERE created_at >= DATE '2026-07-01';

-- On a partitioned table: CREATE INDEX ... ON ONLY the parent, build each partition's
-- index CONCURRENTLY, then ALTER INDEX ... ATTACH PARTITION for each one.

-- Verify: expect an Index Only Scan on idx_orders_recent_status and low shared read buffers
EXPLAIN (ANALYZE, BUFFERS, VERBOSE)
SELECT customer_id, SUM(total_amount) AS revenue
FROM orders
WHERE created_at >= DATE '2026-09-01'
  AND created_at <  DATE '2026-10-01'
  AND status = 'shipped'
GROUP BY customer_id
ORDER BY revenue DESC
LIMIT 100;

T-SQL · SQL Server: availability group, Query Store and health check

-- SQL Server 2022+: availability group with two synchronous replicas and one asynchronous DR replica
CREATE AVAILABILITY GROUP ag_appcore_prod
WITH (AUTOMATED_BACKUP_PREFERENCE = SECONDARY, DB_FAILOVER = ON, CLUSTER_TYPE = WSFC)
FOR DATABASE appcore
REPLICA ON
  N'SQL-NODE-A' WITH (ENDPOINT_URL = N'TCP://sql-a.corp.internal:5022',
      AVAILABILITY_MODE = SYNCHRONOUS_COMMIT,  FAILOVER_MODE = AUTOMATIC,
      SEEDING_MODE = AUTOMATIC, SECONDARY_ROLE (ALLOW_CONNECTIONS = READ_ONLY)),
  N'SQL-NODE-B' WITH (ENDPOINT_URL = N'TCP://sql-b.corp.internal:5022',
      AVAILABILITY_MODE = SYNCHRONOUS_COMMIT,  FAILOVER_MODE = AUTOMATIC,
      SEEDING_MODE = AUTOMATIC, SECONDARY_ROLE (ALLOW_CONNECTIONS = READ_ONLY)),
  N'SQL-NODE-C' WITH (ENDPOINT_URL = N'TCP://sql-c.dr.internal:5022',
      AVAILABILITY_MODE = ASYNCHRONOUS_COMMIT, FAILOVER_MODE = MANUAL,
      SEEDING_MODE = AUTOMATIC, SECONDARY_ROLE (ALLOW_CONNECTIONS = READ_ONLY));

-- Query Store sized so it stays READ_WRITE during incidents (on by default for new databases since 2022)
ALTER DATABASE appcore SET QUERY_STORE = ON
  (OPERATION_MODE = READ_WRITE, INTERVAL_LENGTH_MINUTES = 15,
   MAX_STORAGE_SIZE_MB = 4096, QUERY_CAPTURE_MODE = AUTO);

-- Verify replica health
SELECT ar.replica_server_name, ars.role_desc, ars.synchronization_health_desc
FROM sys.dm_hadr_availability_replica_states AS ars
JOIN sys.availability_replicas               AS ar ON ar.replica_id = ars.replica_id;

Config + SQL · MySQL 8.4: single-primary Group Replication

# my.cnf on every member (MySQL 8.4 LTS); ${...} values are rendered by config management
[mysqld]
server_id                             = 1                  # unique per member
gtid_mode                             = ON
enforce_gtid_consistency              = ON
plugin_load_add                       = group_replication.so
group_replication_group_name          = ${GR_GROUP_UUID}
group_replication_local_address       = 10.0.1.11:33061
group_replication_group_seeds         = 10.0.1.11:33061,10.0.1.12:33061,10.0.1.13:33061
group_replication_start_on_boot       = OFF
group_replication_bootstrap_group     = OFF
group_replication_single_primary_mode = ON

-- First member only, once: bootstrap the group
SET GLOBAL group_replication_bootstrap_group = ON;
START GROUP_REPLICATION USER = '${GR_RECOVERY_USER}', PASSWORD = '${GR_RECOVERY_PASSWORD}';
SET GLOBAL group_replication_bootstrap_group = OFF;

-- Every other member: join with the recovery account (placeholders, never literal secrets)
START GROUP_REPLICATION USER = '${GR_RECOVERY_USER}', PASSWORD = '${GR_RECOVERY_PASSWORD}';

-- Verify: three members ONLINE, exactly one PRIMARY
SELECT MEMBER_HOST, MEMBER_PORT, MEMBER_STATE, MEMBER_ROLE, MEMBER_VERSION
FROM performance_schema.replication_group_members;

Group Replication in single-primary mode gives MySQL automated primary election with majority-certified commits. Consultative Support pairs it with MySQL Router or ProxySQL so clients follow the primary through an election. Engine-level depth continues on our PostgreSQL support, MySQL support and SQL Server support pages.

Backups follow the same discipline: WAL-G or pgBackRest for PostgreSQL, XtraBackup or MariaBackup for the MySQL family, native backup to URL for SQL Server, backint for SAP HANA and clickhouse-backup for ClickHouse, stored in object storage with object lock so a compromised host cannot rewrite the backup chain.

A backup that has never been restored is a hypothesis. Every consultative support engagement schedules quarterly restore drills, measures the recovery time and recovery point actually achieved, and reports the gap against the agreed objective.