search index and jndi

[Bear Giles]

I’m having problems with files (and a derby instance?) for search
indexes even when I’m setting up a completely database-driven backend.
Am I missing something?

Details:

I’m running a recent version of jackrabbit and have set up my
repository.xml file so that all FileSystems and PersistenceManagers,
including the optional FS specified in the <SearchIndex> stanzas, are
actually o.a.j.core.fs.db.DbFileSystem and
o.a.j.core.persistence.bundle.PostgreSQLPersistenceManager,
respectively. I know the database configuration is good because all of
the tables, including the search index tables, are created on a clean
system. (The specifics are tomcat6 server publishing a postgresql
datasource via jndi. Obviously I added the necessary stanza to the
web.xml file as well.)

The problem is that I also see all of the derby(?) stuff under
${rep.home}/index and ${wsp.home}/index. There’s only a single entry the
search
index fsentry table even after populating the workspace.

Am I missing something, some parameter to tell the search index to use
the specified FS?

Thanks,

Bear

(p.s., there seems to be something wrong with the Apache spam filter.
I've been unable to get a message through from my work account, and the
only theory we have here is that the company was acquired a few years
ago and there's probably a disconnect between the domain of the sender
and how the mail server identifies itself. We can't be sure, though,
just that every message gets a spam score of exactly 6.6, which exceeds
the threshold #SMTP# (sic)).

[Bear Giles]

I've documented how I set up my server at
http://wiki.apache.org/jackrabbit/JackRabbitOnTomcat6withJndiAndPostgresql.

Well, one small difference. Before I set up the repository with the
direct PostgreSQL references, but rebuilding it after changing custom
node types I've kept the JNDI configuration. That is, I'm nuking the
database and the directories under /srv/jackrabbit each time and
rebuilding the repository with the JNDI configuration. But as mentioned
earlier it -is- recreating the database tables so I know that the JNDI
setup is correct.

Bear

(p.s., sorry about forgetting to set the 'trivial change' flag on the
small edits!)

Post by Bear Giles
I’m having problems with files (and a derby instance?) for search
indexes even when I’m setting up a completely database-driven backend.
Am I missing something?

[Alexander Klimetschek]

Post by Bear Giles
The problem is that I also see all of the derby(?) stuff under
${rep.home}/index and ${wsp.home}/index. There’s only a single entry the
search
index fsentry table even after populating the workspace.

AFAIK this part of the lucene search index requires fast random access
and that is too slow with the DbFileSystem or is simply impossible to
abstract, because Lucene access the os filesystem directly. But I
could be wrong here....

Regards,
Alex