[DERBY-6943] Undefined behavior of SYSCS_DIAG.ERROR_LOG_READER() when using rolling log files

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 10.13.1.1
• Component/s: Services
• Urgency: Normal
• Assignee: Unassigned
• Reporter: Richard N. Hillegas
• Created: 29/Jun/17 03:03
• Updated: 30/Jun/17 02:10

Description

The behavior of SYSCS_DIAG.ERROR_LOG_READER() is undefined when the engine is booted with -Dderby.stream.error.style=rollingFile. At a minimum, it should scan at least one of the rolling files. Instead, it raises an exception saying that it can't find derby.log.

The following script shows this behavior:

connect 'jdbc:derby:memory:db;create=true';

-- correctly succeeds if using ordinary diagnostic logging
-- INCORRECTLY RAISES 'No such file or directory' if -Dderby.stream.error.style=rollingFile is set
SELECT count(*) FROM TABLE (SYSCS_DIAG.ERROR_LOG_READER()) AS T1;

-- correctly succeeds if using ordinary diagnostic logging
-- correctly raises 'No such file or directory' if -Dderby.stream.error.style=rollingFile is set
SELECT count(*) FROM TABLE (SYSCS_DIAG.ERROR_LOG_READER('derby.log')) AS T1;

-- correctly raises 'No such file or directory' if using ordinary logging
-- correctly succeeds if -Dderby.stream.error.style=rollingFile is set
SELECT count(*) FROM TABLE (SYSCS_DIAG.ERROR_LOG_READER('derby-0.log')) AS T1;

We should decide on a correct behavior for SYSCS_DIAG.ERROR_LOG_READER() when the -Dderby.stream.error.style=rollingFile property is set. Here are some options:

1) Scan all of the rolling diagnostic log files, starting at derby-0.log and
continuing through the last in the series.

2) Scan just the first rolling diagnostic log file.

3) Scan just the last (current) diagnostic log file.