pg.ddx.io  aws/aws-advanced-jdbc-wrapper GitHub issues and pull requests  
help / color / mirror / Atom feed
From: aaron-congo (@aaron-congo) <noreply+aaron-congo@github.com>
To: aws/aws-advanced-jdbc-wrapper <noreply+aws-aws-advanced-jdbc-wrapper@github.com>
Subject: Re: [aws/aws-advanced-jdbc-wrapper] PR #2237: feat: add efm3 host monitoring plugin with per-connection registration
Date: Wed, 18 Nov 2026 22:07:31 +0000
Message-ID: <gh-aws-aws-advanced-jdbc-wrapper-2237-rc4621515480@github.com> (raw)
In-Reply-To: <gh-aws-aws-advanced-jdbc-wrapper-2237@github.com>
References: <gh-aws-aws-advanced-jdbc-wrapper-2237@github.com>

(on wrapper/src/main/java/software/amazon/jdbc/plugin/efm/v3/HostMonitoringConnectionPluginV3.java)

A slot that was claimed `ABORTED` while idle stays `ABORTED` until its next call. That call then marks the host `NOT_AVAILABLE`, no matter how long ago the abort happened or whether the host has recovered since.

Example with `abortIdleConnectionsOnFailure=true`: a short blip marks the host unhealthy, and idle connections are claimed and aborted. The host recovers within seconds. Minutes or hours later the app uses one of those pooled connections, `enter()` returns true, and the plugin marks a healthy host unavailable. That steers host selection, read/write splitting, and failover away from it.

Could this check whether the monitor still considers the host unhealthy (`monitor.isHostUnhealthy()`), or record when the abort happened and ignore old ones? The call fails on the dead connection either way. This only affects whether a stale availability signal is published.


view thread (39+ messages)  latest in thread

Message-ID: <gh-aws-aws-advanced-jdbc-wrapper-2237-rc4621515480@github.com>
Permalink:  ../gh-aws-aws-advanced-jdbc-wrapper-2237-rc4621515480@github.com/

 · 


This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox