8.3.0+: Pre-seeding cluster.properties causes Jira Software to fail to start

Issue Summary

Pre-seeding cluster.properties in a new install causes Jira to fail to start.

I have marked this critical as it affects the Jira quickstarts, which rely on pre-seeding the database and clustering.

Environment

Linux (confirmed on Amazon Linux 2 on EC2, and Ubuntu 18.04 in VirtualBox).

Steps to Reproduce

• Install Ubuntu 18.04 on a local VM.
• Install packages postgresql-10 and openjdk-8-jre
• Setup Postgres, add a blank database jira
• Download atlassian-jira-software-8.3.1.tar.gz and unpack
• Create ~/jira/ and ~/jira-shared/
• Create the files ~/jira/dbconfig.xml and ~/jira/cluster.properties (see attached for versions I used).
• export JIRA_HOME=~/jira/
• Start Jira with ./bin/start-jira.sh -fg

Connecting to http://localhost:8080 hangs indefinitely.

The logs contain a lot log lines like:

2019-08-05 04:16:22,940 http-nio-8080-exec-2 WARN anonymous 256x3x1 35c7u6 10.0.2.2 /secure/SetupDatabase!default.jspa [n.sf.ehcache.Cache] Data availability impacted:
****************************************************************************************
************************** removeAll called on a pinned cache **************************
****************************************************************************************

Additionally, the following lines appear, however they also appear in a clean install (no-pre-seeding):

java.lang.NoSuchMethodException: org.apache.commons.dbcp2.PoolingDataSource$PoolGuardConnectionWrapper.getPrepareThreshold()

Notes

• This occurs on both Software and ServiceDesk, but not on Core.
• Seeding only dbconfig.xml does not seem to produce the error.
• Two logs have been attached, one from a clean (unseeded) install, and from a run seeded with both dbconfig.xml and cluster.properties.

Workaround

Currently there is no known workaround for this behavior. A workaround will be added here when available