portupgrade and index
list-freebsd-2004 at morbius.sent.com
Tue Nov 30 05:23:50 PST 2004
On Tuesday 30 November 2004 05:58, Kent Stewart wrote:
> On Monday 29 November 2004 09:40 pm, Kris Kennaway wrote:
> > On Mon, Nov 29, 2004 at 11:27:47PM -0600, Conrad J. Sabatier wrote:
> > > On Mon, 29 Nov 2004 19:57:34 -0800, Kris Kennaway
> > > <kris at obsecurity.org>
> > >
> > > wrote:
> > > > On Mon, Nov 29, 2004 at 10:48:29PM -0500, dave wrote:
> > > > > Hello,
> > > > > I've got a box that runs portupgrade to keep the most
> > > > > current ports.
> > > > > I've lately switched to make fetchindex vs. portsdb -uU which
> > > > > goes much faster. My problem is whenever i do a make search for
> > > > > a port the index.5 file is regenerated and that takes an
> > > > > extremely long time. I was wondering if this is normal
> > > > > behavior, and what if anything i could do to speed it up?
> > > >
> > > > make fetchindex
> > > >
> > > > Kris
> > >
> > > You can also increase the concurrency of "portsdb -U"/"make index"
> > > by setting INDEX_JOBS in /etc/make.conf. The default number of
> > > parallel jobs is 2. Increasing this to, say, 8, will save some
> > > time in building the index.
> > In my testing 4 helped on a dual SMP machine but 8 didn't, because
> > the process was I/O bound already at 4.
> There isn't a "make index" that can compete with downloading an
> INDEX.bz2. I timed a make fetchindex and it required all of 11 seconds
> on my DSL line. That would work out to around 3 minutes on a dial up.
I don't think anyone in this sub-thread has read the original question
correctly. As I understand it he saying that "make search" generates a new
index despite having already run "make fetchindex".
More information about the freebsd-questions