MemStore pressure means the in-memory write buffer for a tenant is near capacity. Identify the tenant with memstore_percent near 100, confirm whether minor freezes are completing, reduce write burst or expand resources, and escalate if freezes stall repeatedly.
Symptom
Writes stall or latency spikes; alerts on memstore usage for one tenant.
Impact
Tenant-scoped write outage or severe slowdown until memstore pressure clears.
Preconditions
- OCP or SQL access to tenant memstore metrics.
- Authority to pause batch jobs affecting the tenant.
Diagnostic commands (read-only)
Query tenant memstore usage via OCP or virtual tables documented for your OceanBase 4.x build. Compare with compaction and freeze metrics on the same observer.
Decision points
| Observation | Action |
|---|---|
| Transient spike | Monitor; freezes may self-resolve |
| Sustained 100% | Reduce writes or expand unit memory |
| Freeze stuck | Escalate; possible compaction or disk issue |
Remediation
- Pause non-critical bulk loaders after stakeholder approval.
- Schedule major compaction in maintenance window if backlog is structural.
- Avoid ad-hoc memory changes without change control.
Validation
memstore_percent below your operational threshold for 15+ minutes. Write latency returns to baseline.
Escalation
Repeated incidents on same tenant within 24 hours. Suspected compaction failure or disk errors.