Blue-Green-Deployment-Plugin Questions

[dvdgsng]

Hi, we're testing the BGD functionality for our Keycloak & Infinipan clusters. The BGD-Switchover itself works, the workloads recovered the db connections as soon as the new cluster topology was available and used the green-now-blue db instances. But a few questions arose...

Env

• Aurora PostgreSQL
• PostgreSQL Engine: 15.10
• Connection String:

    driver: "software.amazon.jdbc.Driver"
    protocol: "jdbc:aws-wrapper:postgresql"
    config: >-
      wrapperPlugins=initialConnection,auroraConnectionTracker,failover2,efm2,bg
      &failoverMode=strict-writer
      &enableGreenNodeReplacement=true
      &wrapperLoggerLevel=INFO
      &enableTelemetry=true
      &telemetryMetricsBackend=OTLP
      &telemetryTracesBackend=OTLP

• loglevel:

      - category: software.amazon.jdbc
        level: INFO
      - category: software.amazon.jdbc.plugin.bluegreen.BlueGreenStatusProvider
        level: ALL

1. BGD Detection & Status

Detection of a BGD does not seem to work for us. Log output is the same before, during and after a BGD exists. We've never seen a logSwitchoverFinalSummary even though switch-over is successful and applications recover successfully. So, what is the expected log output of the plugin for the BGD lifecycle (created, replicating, switching-over, deleted)?

Some log details:

• Status Phase is always NOT_CREATED

[bgdId: '1'] Summary status:
software.amazon.jdbc.plugin.bluegreen.BlueGreenStatus@427fb4b9 [
 bgdId: '1', 
 phase: NOT_CREATED, 
 Connect routing:    - 
 Execute routing:    - 
 roleByHost:    - 
]

• Phase time is always someting like this

Phase times:
   NOT_CREATED -> 2025-08-12T12:24:14.036897053Z

• Corresponding nodes are always empty

• Green node certificate change times is always empty

• Interim State always the same:

[bgdId: '1', role: SOURCE] software.amazon.jdbc.plugin.bluegreen.BlueGreenInterimStatus@70e5c8bc [
 phase NOT_CREATED, 
 version '1.0', 
 port -1, 
 hostNames:  - 
 Start Topology: 
 start IP map:   - 
 Current Topology:  
 current IP map:   - 
 allStartTopologyIpChanged: false 
 allStartTopologyEndpointsRemoved: false 
 allTopologyChanged: false 
]

• This remains the same too:

  latestStatusPhase: NOT_CREATED
   blueDnsUpdateCompleted: false
   greenDnsRemoved: false
   greenNodeChangedName: false
   greenTopologyChanged: false

• During Switchover the log output changes shortly to just "[bgdId: '1'] BG status: NOT_CREATED" ... unfort. no progress details at all are printed on log level info.

2. bgdId

What exactly is the bgdId? Is it some kind of rds resource name (eg the name of the BGD ("bg-deployment-1") or the BGD Resource ID ("db-BTEVE7M4DEPI6BLQZQRPSM54RE")? Or a custom string only used by the wrapper internally (e.g. "my-bgd-42"). Is it used by the plugin to detect the BGD at all? We've tried to set different values without changed results.

3. enableGreenNodeReplacement

Should the previous parameter "enableGreenNodeReplacement" be removed when using the bg-plugin? Does it interfere with the bg-plugin?

4. PostgreSQL log spam

Missing function creates heavy log output on the db instances:

keycloak:[31100]:ERROR: function "get_blue_green_fast_switchover_metadata" does not exist at character 8
keycloak:[31100]:STATEMENT: SELECT 'get_blue_green_fast_switchover_metadata'::regproc

Is there any way to reduce this?

---

Thanks in advance! :)

[hsuamz]

Hello @dvdgsng,

We've updated our documentation here: https://github.com/aws/aws-advanced-jdbc-wrapper?tab=readme-ov-file#amazon-rds-bluegreen-deployments.

For visibility, we've been working with the service team in regards to the metadata and the metadata is currently not available in the current RDS and Aurora service. Please contact your AWS account team for metadata release timelines.

Once the metadata is available on the service, the JDBC driver will support Blue/Green deployments.

Thanks,

Matthew