Error during upgrade to 7.11.1 (from 6.15.2): ERROR: relation "confzdu" does not exist

[Karl Romike]

The forum suggested bumping up the -XMS / -XMX parameters in setenv.sh.  In two separate tests (1024m -> 2048m) and (2048m -> 4096m), the result is the same.

After issuing a 'service confluence start' on the sandbox host, the last entry in catalina.out starts with:
2021-03-18 21:07:31,051 WARN [Catalina-utility-1] "[confluence.impl.setup.DefaultBootstrapDatabaseAccessor] readValue Unable to determine ZDU status from database. If you are upgrading from a Confluence version prior to 7.9, this is expected.
org.postgresql.util.PSQLException: ERROR: relation "confzdu" does not exist...
...Security framework of XStream not explicitly initialized, using predefined black list on your own risk."

The back-end database is an AWS RDS instance running PostgreSQL 9.6.

The only occasional log activity is the gc*.current.  A netstat shows that 8090 is listening, but a Chrome browser session just spins w/o ever timing out.  One of the posts stated that the upgrade takes longer than usual and to "just wait."  I waited over 30 minutes before deciding to submit this inquiry.

A rollback to 6.15.2 brings everything back up within 5 minutes.

Thank you.

Karl Romike

SkyTouch Technology

Phoenix Az.

[Karl Romike · Accepted solution]

Question withdrawn; solution discovered by accident. 

What the online documentation doesn't readily tell you (at least that I could find within the boundaries of my paper-thin patience) is that if you take the manual approach (un-tar the binaries, update the confluence-init.properties and bounce the service), the Confluence app then kicks off a 100%-background upgrade of the RDS db-database. 

While this now seems obvious in hindsight, no logging (e.g. logs/catalina.YYYY-MM-DD.log) mentions this.  The service appears to simply stall after the line "INFO [Catalina-utility-2] org.apache.catalina.core.ApplicationContext.log Initializing Spring DispatcherServlet 'dispatcher'".

Poking around in the RDS database instance CW metrics, I saw that there was Read/Write IOPS activity, despite this being a solitary sandbox with no browser login page.  Approximately 1 hour & 45 minutes later, the db upgrade completed and the service (via catalina.log) completed its startup.

So, I am eating some crow in deference to a post where the author correctly stated "it takes awhile".  I just wish he or she had defined "awhile" for better context.