On Fri, 2011-05-06 at 08:00 -0700, John Reiser wrote:
> On 05/04/2011 08:17 AM, David Lehman wrote:
> > On Tue, 2011-05-03 at 22:23 -0700, John Reiser wrote:
>
> >> One guess for needing so much RAM is that the installer's allocator
> >> (python/C memory manager) never reclaims anything unless forced.
> >> The amount of memory _apparently_ required increases package-by-package.
> >> [snip]
>
> > A nice theory, but it shows that you have no understanding of how
> > package installation works. [snip]
The above comment was with regard to your prescribed solution involving
installing each package in a separate process, not the notion that large
rpm/yum transactions have a large memory footprint.
<snip>
> So anaconda's heap grew monotonically by 27MB over 1205 packages,
> which is an average of 22kB per package. My guess was spot-on.
> Also, the heap grows about 100MB for "installation starting".
> Both amounts exceed 1% of 1GB system RAM with no swap space,
> and deserve further investigation.
It has been known anecdotally for some time that dependency checking and
initial transaction setup have memory requirements roughly proportional
to the size of the package set, and that overall memory requirements for
processing the transaction are also roughly proportional to the number
of packages being installed. As you might imagine, this is almost all,
if not all, inside yum and rpm. Feel free to send mail to the
appropriate mailing lists if you wish to work with the maintainers of
those packages to improve the situation.
Tailoring the installation process to minimally-configured systems is
not at the top of our priority list, but it is on the list. Your
patience is appreciated.
Dave
>
>
> -----ps-smaps.sh
> mkdir -p $1
> cd $1
> ps l >ps.out
> for pid in $(awk '{print $3}' < ps.out); do
> cp /proc/$pid/smaps $pid.smaps 2>/dev/null
> done
> ----- top-level during package install
> while true; do
> date=$(date "+%H%M%S")
> sh ps-smaps.sh /tmp/smaps/$date
> sleep 60
> done
> -----
>
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
