3f76bc9507
Some checks failed
Deploy to Phoenix / deploy (push) Has been cancelled
- MASTER_INDEX: Last Updated 2026-03-06; status 59/59 contracts; add NEXT_STEPS_LIST, CONTRACT_NEXT_STEPS_LIST - docs/README, NEXT_STEPS_INDEX, 06-besu/MASTER_INDEX: Last Updated 2026-03-06 - Contract check script: 59 addresses (PMM, vault/reserve, CompliantFiatTokens); canonical CCIP/router - New docs: EXECUTION_CHECKLIST, NEXT_STEPS_LIST, DOTENV_AUDIT, ADDITIONAL_PATHS, deployer gas runbook, WEMIX_ACQUISITION_TABLED, etc. - Config: deployer-gas-routes, cro-wemix-swap-routes, routing-registry, token-mapping - Scripts: check-contracts-on-chain-138, check-pmm-pool-balances-chain138, deployer-gas-auto-route, acquire-cro-and-wemix-gas - Operator rule: operator-lan-access-check.mdc Made-with: Cursor
1.8 KiB
1.8 KiB
DBIS PostgreSQL HA Pair Runbook
Topology
- Primary writer:
192.168.11.105(CT10100) - Replica reader:
192.168.11.106(CT10101) - Proxmox host:
192.168.11.11 - Database:
dbis_core
Why this matters
- Writes must go to exactly one primary.
- Reads/indexing should prefer replica.
- If both nodes are writable, split-brain risk is high.
1) Convert to true HA (streaming replica)
Use once to enforce primary/replica:
cd /home/intlc/projects/proxmox
REPL_PASSWORD='<replication-password>' \
scripts/database/convert-dbis-postgres-to-ha.sh
2) Verify HA health
cd /home/intlc/projects/proxmox
DB_PASSWORD='<db-password>' \
scripts/database/check-dbis-postgres-ha.sh
Expected:
192.168.11.105:pg_is_in_recovery = f192.168.11.106:pg_is_in_recovery = t- Replica write test fails with read-only transaction error.
3) Correct connection policy
- Application writes:
192.168.11.105:5432 - Read-only APIs/indexers:
192.168.11.106:5432 - Admin checks:
- Primary replication status:
sudo -u postgres psql -Atc "select client_addr,state,sync_state from pg_stat_replication;"
- Replica recovery status:
sudo -u postgres psql -Atc "select pg_is_in_recovery();"
- Primary replication status:
4) Explorer label/indexing publishing
- Publish labels/tags to primary only (
192.168.11.105). - Replica receives updates via WAL streaming automatically.
- Do not run dual-write mode when streaming replication is healthy.
5) Optional hardening
- Add VIP (keepalived/HAProxy) for stable writer endpoint.
- Enable synchronous replication if zero-data-loss failover is required.
- Add scheduled health checks + failover playbook.
6) Notes on rsync
- Do not use rsync for live PostgreSQL data directories.
- Use streaming replication for live sync.
- Use rsync only for ancillary files/runbooks/backups, not active PG data.