Last updated Jun 14, 2026

Partition split (SEV2)

Triggered by: a hot table's current partition exceeds size threshold (typically 50M rows / 100GB). apps/api/lib/partition-splitter from P0.H9 normally handles this automatically; manual intervention only on degenerate cases.

Diagnose

matter ops partition-status --table <name>

Identify:

• Current partition size + row count.
• Next-partition readiness (precreated?).
• Recent write rate per partition.

Auto-split path

Normally the rolling-create cron handles partitioning:

matter ops partition-create-next --table <name>

This is the standard daily path.

Manual intervention

When auto-split degenerates:

1. Plan target boundary (typically monthly: month-of-creation).
2. Pre-create partitions for the next 12 months if catching up.
3. Confirm EXPLAIN plans still prune correctly post-split:

   matter ops verify-query-plan --table <name>

Validate

matter ops verify-partitions --table <name>

Expected: every partition under threshold, EXPLAIN plans prune to single partition for time-bounded queries.

Post-recovery

1. Adjust auto-split threshold if the default tripped late.
2. Verify hot-row table list is current.