GBase 8a MPP Cluster provides three operating modes that control what SQL operations are allowed: NORMAL, READONLY, and RECOVERY. Administrators switch between them manually to match the current operational need — from everyday production to maintenance windows and disaster recovery.
The Three Modes at a Glance
| Mode | Allowed SQL | Forbidden SQL | Purpose |
|---|---|---|---|
| NORMAL | Everything: SELECT, DML, DDL, LOAD DATA | Nothing | Day‑to‑day production with full read/write capability. |
| READONLY | SELECT queries only | All writes: DML, DDL, LOAD DATA | Safe maintenance window — freezes data while keeping queries running. |
| RECOVERY | Nothing (not even SELECT) | Everything | Extreme data protection during restore or critical repair. |
When to Use Each Mode
NORMAL: Default Production Mode
The cluster serves all reads and writes. Major maintenance that could affect consistency or availability should not be performed in this mode.
READONLY: Standard Maintenance & Change Window
This is the go‑to mode for planned operations — effectively a "maintenance mode". Use it for:
- Scaling up/down: Prevents concurrent writes during data redistribution.
- Node replacement: The cluster can automatically or manually enter read‑only mode when swapping a failed node.
- Backups: Ensures a consistent data view for logical backups.
- Major application releases: Briefly freeze data while deploying.
RECOVERY: High‑Risk Repair Mode
Reserved for specific recovery scenarios and used only by senior administrators:
- Restoring a cluster or VC from backups.
- Disaster recovery drills.
- Fixing severe logical corruption by importing external repair data.
All connections and queries are rejected in this mode — business is completely interrupted.
Viewing and Switching Modes
Check the Current Mode
gcadmin showcluster vc vc1
Look for the VIRTUAL CLUSTER MODE field in the output.
Switch Modes Manually
# Set the entire cluster to read‑only
gcadmin switchmode readonly
# Set a specific VC to recovery mode
gcadmin switchmode recovery vc vc1
# Return to normal mode
gcadmin switchmode normal
Best Practices
| Mode | Frequency | Business Impact | Rule of Thumb |
|---|---|---|---|
| NORMAL | Always on | None | Production runs with full features. |
| READONLY | Often (maintenance) | Queries unaffected; writes paused | Switch to read‑only before any change. |
| RECOVERY | Rarely | Full outage | Use recovery only at the foundation level. |
- Always switch to
READONLYbefore planned maintenance such as node changes or backups. -
RECOVERYmode is an "operating room" — notify stakeholders and have a rollback plan. - Mode switches are fast, but schedule them during off‑peak hours when possible.
Using these three modes wisely keeps your gbase database safe and your maintenance operations smooth. It's a simple but powerful part of the GBASE cluster management toolkit.