## Improvements

### Improve Health Checker

The health checker has been modified to reduce the number of unnecessary fail-overs. Changes to the health checker include:

- The PostgreSQL health-check makes fewer calls to PostgreSQL
- All PostgreSQL failures are subject to health check retries (bug fix)
- Removed the etcd health-check because the main leader election detects etcd failures first

### ionice Configuration Options

New `ionice` settings for controlling the disk input/output scheduling `class` and `set` priority for the `etcd` process
have been added. `etcd.ionice.class` and `etcd.ionice.level` settings with the default values of `2` and `0` respectively. The default values mirror the current system defaults on Linux.

### Leader Endpoint Caching

The `/leader` endpoint now serves a locally cached response. This should reduce the possibility of timeouts returned to clients polling the `/leader` endpoint. The Chef Infra Server polls the `/leader` endpoint, and thus we believe this change will reduce some Chef Backend outages that resulted from `/leader` request timeouts.

## Bug Fixes

- Fixed PostgreSQL validator problem that prevented `join-cluster` from running to completion

## Component Updates

### Chef Infra Client 15.12.22

The embedded Chef Infra Client used for Chef Backend configuration has been updated from 14.14.25 to 15.12.22

### Ruby 2.6.5

Ruby has been updated from 2.5.7 to 2.6.5 for improved performance.