Re: [Tracker] tracker on a desktop cluster?

["Chris Fanning" <christopher fanning gmail com>]

On Tue, Nov 11, 2008 at 10:24 AM, Martyn Russell <martyn imendio com> wrote:
> Chris Fanning wrote:
> > Hi,
>
> Hi :)

Ok, I tried it and it works. Users are going to love this!

I think it would be nice to be able to pass a different
/path/to/my/index_server/tracker.cfg to trackerd on the index_server,
or equally be able to do the same when starting trackerd on the node,
but I hacen't seen an option for this in the manpages.

One small problem.
I start trackerd on the index_server. And when I start
"/usr/bin/trackerd --no-indexing" on the node it reports "WARNING **:
Tracker daemon is already running - exiting"
So I must kill trackerd on the index_server first. Is there anyway to
get round this?

Thanks.
Chris.

> > I'm new on the list.
> > I'd like to know if what I'm planning with tracker is possible or what
> > might be the way to get what I want working.
> >
> > I have a small cluster that serves ubuntu desktops to users via
> > freenx. I would like to make tracker-search-tool available to the
> > users.
> > But I'm not sure how to do it. I've read quite often on the net about
> > tracker running wild or using lots of system resources.
> > This is something I must avoid because on each node there can be up to
> > 15 concurrent users.
> > So I thought I could set up a server solely dedicated to indexing
> > files. This index_server (lenny?) could import users homes via NFS.
> > I could then write some kind of script that runs something like for each user:
> > su - uid
> > dbus-launch trackerd
> >
> > Then back on the node the user would run
> > trackerd  --no-indexing
> >
> > Would this work?
>
> It should do.
>
> > What will happen when the nodes and the index_server
> > are running different versions of tracker?
>
> We have a database version file which is checked for this. Right now it
> will reindex if the version is older than the current version installed.
> Of course, if you pass --no-index, it will reset everything and have no
> data until you run trackerd normally.
>
> > Might there be another way to deal with this scenario?
>
> Well, this situation is a little untested. We have mechanisms in place
> to try to work round this.
>
> --
> Regards,
> Martyn