What are the conditions that cause corruption? If there is just one
writer and multiple readers, is that safe?
The cases are well spelled out in Lucene in Action, section 2.9.
Generally, one writer and multiple readers is not safe for disabling
locking.
For example, the IndexReader, when it is first instantiated, needs to
read segment data structures from the index. If an IndexWriter happens
to be committing at the same time, this will cause corruption. Also, if
an IndexReader is doing deletes, an IndexWriter is adding new docs, that
could also cause corruption (if you turn off locking).
Unless you have your own locking layer over Lucene it's generally
dangerous to turn off its default locking.
Also: you should generally try to have a single IndexReader (and
IndexSearcher using it) that's shared across all "things" that need to
run searches.
One question: why do you want / need to disable Lucene's default
locking? I'm trying to work through refactoring how Lucene does locking
(see LUCENE-305 in Jira) so details on your use case can help guide
these fixes.
Mike
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
