Re: [Help-gnunet] help with database management

[Igor Wronsky]

On Sun, 12 Oct 2003, Benjamin Paul Kay wrote:

> > Hmm. DATABASEBUCKETS is no longer used. Does current 'gnunet-check -a'
> > think that all is fine or does it complain about something?
> gnunet-check -a reported everything is just peachy. It also reported only one
> file in the database and took all of fifteen seconds to run. Somehow I think
> it should take a bit longer to check the entire database...

Right. Somehow, its not putting the content to the db properly
I suppose. But then, you shouldn't be able to download them yourself
either. Odd.

> Should I remove the line "DATABASEBUCKETS = 1" from gnunet.conf? If so, do I
> need to run gnunet-convert afterwards?

If you're already running 0.6.x now, no. You can just drop the line.

> > Are all the indexed files listed properly in
> > $GNUNET_HOME/data/afs/database.list
> The file exists but is empty.

Thats bad, unless you've specified '-n'. Even then, the check should
be slower.

> > (type it)? You might also want to look at the contents
> > of the dir where MySQL holds gnunets tables and check that
> > there's nothing suspicious there (all *.MYD files should
> > be roughly the same size) and perhaps pull a 'repair table'
> > on each of them (when gnunetd is not running).
> I'm new to mysql. Could you elaborate?

Shutdown gnunetd. In shell, do

$ mysql -u username
mysql> use gnunet;
mysql> show tables;
mysql> repair table data1024of0;
mysql> <and same for the rest of the tables>

(naturally use the names for the tables that 'show tables' reports)

mysql> quit
$ gnunet-check -a

and restart gnunetd. Though this will make sure that the mysql
database is structurally intact and that gnunet-check can't find
obvious problems with it, I'm not so sure it will fix this
particular problem. But in any case, you'll have to reindex
the content. It must appear to database.list or just
to the db, if you use full insertion.

> Will do. Its a shame there isn't a URI indexing service like ShareReactor for
> GNUnet

I agree. If some third party from Honolulu (or some other place in the
free world) wants to put up such a service, that'd be nice, but the devel
team members living in oppressive regimes can't participate in such an
effort. :(

> together on search efficacy (no offense). Perhaps a URI could be posted on
> GNUnet's homepage that points to a namespace with an interval-updated list of
> URI's of indexed files in it... just an idea.

Or somebody can put up a site with such an URI and people can find
the URI with Google. ;)


Igor