AngaraReplica v2 Operations Guide

Short operator guide for streaming replication v2. Canonical source: this runbook in angarabook/src/operations/.

Topology and scope

• 1 primary + up to 8 standby nodes (async replication).
• Standby works in read-only mode (SQLSTATE 25006 on write).
• Promote is performed manually (auto-failover in the next major line).

Configuration baseline

Primary:

• [replication].role = "primary"
• listen_addr
• wal_retention_segments

Standby:

• [replication].role = "standby"
• primary_addr
• slot_name
• wal_path

Operations flow

1. Start primary.
2. Start standby and check lag metrics.
3. Monitor replication lag / reconnects / slots.

Promote (manual failover)

• Promote must complete through the sync-checkpoint handshake.
• Promote timeout fails closed (standby does not accept writes if the handshake did not complete).
• Lease-based fencing reduces split-brain risk, but does not fully replace STONITH/Raft.

Key monitoring signals

• angara_node_is_standby
• angara_replication_lag_bytes
• angara_replication_lag_ms
• angara_replication_reconnects_total
• angara_promote_total
• angara_promote_duration_ms_last

Typical incidents

• Standby does not connect: address/port/firewall/reconnects.
• WAL segment gone: base backup and standby restart are required.
• Promote timeout: check network and WAL write path on primary.

Next

• Disaster recovery playbook — DR scenarios on top of replication.
• Backup and restore (operator-level) — how replication complements (does not replace) backup.
• Operational policies baseline — SLA/RTO/RPO agreements within which replication v2.