Solr11035Bandaid failure messages, ReindexCollectionTest.testSameTargetReindexing

[Erick Erickson <er...@gmail.com>]

FYI, “bandaid” is a bit of code I put in to try to understand if we just had timing issues for test failures to find the expected number of documents. 

Seeing messages about bandaid code succeeding means that we, perhaps, failed to commit in time, somehow skipped a commit, the searcher hasn’t warmed etc, but the index has the expected docs.

Seeing messages about baindaid _failure_ means that the index is _not_ in the expected state, and there have _certainly_ been successful commits since the index was modified by any test calling the bandaid code, i.e. it’s not a timing issue.

As of 30-Sep about 12:30 Eastern time, it works like this:

- check the count, if it’s as expected, return
- inset a bogus doc, commit and spin until it’s found.
- remove the bogus doc, commit and spin until it’s _not_ found
- try the original query again and fail if the counts are not the same.

So testSameTargetReindexing is, indeed, producing an index that doesn’t have the counts in the target that were in the source.

And, FWIW, if anyone is using the bandaid code for similar purposes, there’s an optional flag to fail even if the doc count are equal after inserting and deleting the bogus doc, useful when you want to see failures, well, anyway.

Erick


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org