CLUSTER MANAGEMENT

• Upgrading your cluster
  • Before you upgrade
    • Verify cluster health
    • Verify backups
    • Verify replication lag
    • Pin and review package versions
    • Test the upgrade in a staging environment
    • Disable scheduled jobs
  • During the upgrade
    • What to expect
    • What to watch from the application side
  • After the upgrade
    • Verify installed package versions
    • Verify services and replication
    • Take a fresh backup
    • Re-enable scheduled jobs
    • Run tpaexec test
  • Recovering from a failed upgrade
    • When to roll back, when to call EDB Support
  • Common pitfalls
    • Pinned package versions cause no-op upgrades
    • Shared Barman or shared PEM clusters
  • Component selection
  • Package version selection
  • Configuration
  • Upgrading from BDR-Always-ON to PGD-Always-ON
  • Upgrading from PGD-Always-ON to PGD-X
    • Prerequisites
    • Stage 1: Migrating to the Built-in Connection Manager
    • Step 1.2: Switch to Connection Manager
    • Stage 1 Complete
    • Stage 2: Upgrading the Architecture to PGD-X
    • Step 2.2: Perform the Software Upgrade
    • Upgrade Complete
  • PGD-S or PGD-X
  • PGD-Always-ON
  • BDR-Always-ON
  • M1
  • Controlling the upgrade process
  • Upgrading a Subset of Nodes
    • Best Practice for PGD-Always-ON/BDR-Always-ON
    • PGD-S and PGD-X
• tpaexec switchover
  • Example
  • Architecture options
  • Repmgr redirect pgbouncer
• BDR/HAProxy server pool management
  • Example
• tpaexec rehydrate
  • Prerequisites
  • Example
    • Change the configuration
    • Start the rehydration
  • Rehydrate in phases
  • Appendix
    • Using awscli to change volume attributes