FilteredTimeSeries has a failing Integration Test with latest schema

[rma-rripken]

When the Integration Tests run against the latest schema there is a consistent failure in TimeSeriesFilteredControllerIT.

The test is creating a timeseries specified as Calhoun.Flow.Inst.~1Hour.0.cda-test. When the test runs against the latest schema there is this error:
ORA-20998: ERROR: Invalid ID for PRTS: Calhoun.Flow.Inst.1HourLocal.0.cda-test

Notice that ~1Hour became 1HourLocal.

Is 1HourLocal even valid?

For some reason the timeseries is getting created as LRTS instead of PRTS.

The filtered tests should pass in both schemas

[MikeNeilson]

... it is not. it's only ~2HourLocal and longer intervals.

That said it is a bit of an edge case.

Though the test is probably running after a test that enables LRTS mode for the session, probably good to to just set it correctly for the current tests.

And odd that the difference in behavior is between the schema versions. @perrymanmd ?

[rma-rripken]

Oh, that is almost certainly the problem. I didn't realize JooqDao was modified to set the LRTS flag in the session. The way its coded now is the issue. If the ctx flag is set it gets put in the session but then nothing clears it from the session.

We could change that logic to set it in the db session whether its true or false and that would probably fix things.

A possibly better way to do it might be to make a DataSource that conditionally return wrapped Connections that unset the LRTS in the session on Connection close(). The wrapper approach could probably even be conditional - you only get a wrapped connection if use_lrts was true. Then every place that needs db access doesn't pay an lrts tax.

[rma-rripken]

I locally added an else case to JooqDao.getDslContext that logged CWMS_UTIL_PACKAGE.call_GET_SESSION_INFO(dslContext.configuration()
, SESSION_USE_LRTS_ID_FORMAT);

And indeed, when the full suite is run the session already contains the LRTS flag 'T' and num format 6.

When TimeSeriesFilteredControllerTest is run by itself the session does not contain the flag or format value.

[zack-rma]

The LRTS testing was intended to revert the format to the default upon test completion, but it sounds like one or more of those were missed.

I like the idea of wrapped connections, as that would eliminate the possibility of leaving the LRTS format in an undesirable state.

I am curious about the failure only occurring on the latest schema, not sure what would have changed recently to cause that.

[rma-rripken]

I changed the FilteredTimeSeries test to not use ~1Hour. The underlying issue with LRTS was not fixed

[rma-rripken]

Created a new issue to track the JooqDao problem. #1588