hi,

I ended up starting a bulk build a couple times myself, and the last one
ran for about 2 weeks before crashing with:

	panic: malloc: out of space in kmem_map

that was probably due to my using large MFS file systems for /tmp and /var/tmp,
but it did seem that there was more fragmentation of their vm_maps than there
should have been...  looks like the amap-merging code isn't quite working.
I haven't had a chance to follow up on that yet though.

I also got several instances of:

	sd0(esiop0:0:8:0): command timeout, CDB: 0x2a 00 00 30 13 48 00 00 20 00
	esiop0: scsi bus reset
	sd0: async, 8-bit transfers, tagged queueing
	sd0: sync (50.00ns offset 16), 16-bit (40.000MB/s) transfers, tagged queueing

but that didn't seem to cause any problems.  maybe the disk is a little flakey.

the worse problem was that after the kmem_map panic, the file system containing
the pkgsrc tree was pretty badly corrupted... thousands of bad inodes and
eventually fsck dumped core.  I haven't looked into that very much yet.

it did manage to build quite a few pkgs while it was going, though:

# ls packages/All/ | wc
    1378    1378   26989
# wc -l .broken.html 
     481 .broken.html

so at this point I wouldn't recommend bothering with another bulk build
unless you're interested in debugging the kernel problems that it exposes.

-Chuck


On Fri, Aug 26, 2005 at 08:03:04PM -0400, David H. Gutteridge wrote:
> Hi all,
> 
> I noticed back in your May archives there was discussion about finding
> someone to do pkgsrc bulk builds.  Someone had expressed interest, but
> the thread petered out.
> 
> I have a B180L with half a gig of RAM in it sitting idle (and a decent
> internet connection), I can commit it to doing pkgsrc builds if that's
> of use, with one caveat.  I'm going to be indisposed for part of
> September,
> and I'd also like to tinker with the B180L a bit first.  If the
> group
> is interested, what I'd propose is I wait for pkgsrc-2005Q3 then
> do a
> build on that, as it may be more telling to use a branch that's
> stabilized.
> 
> Since I see boot from disk works now, I'm inclined to run things that way,
> and just upgrade the hard drive that's in the machine right now to something
> that'll handle the space requirements of a bulk build (I have a drive
> on-hand
> I can use).  But I thought I should also ask: has anyone run into any stability
> problems over the long-term after booting from disk (problems suspected
> to be related to boot from disk, of course)?
> 
> Regards,
> 
> David Gutteridge
>