Yes, I have a N330 (zotec ION) with 3G ram, no local storage and swap
over nbd with portage and build area in /tmp which itself is on tmpfs.
Some packages (gcc and glibc in particular) require a lot of ram and
tmpfs to emerge so sometimes I have to disable tmpfs and use nfs
storage.
Because portage is lost on reboot when using tmpfs, an emerge sync is
needed to rebuild it when an update is needed and that takes time (more
than just updating portage over nfs in fact). The upside is emerges can
be very fast indeed and updating portage is an unattended operation so
cost (admin time) is small for me.
Only disadvantage is gcc and possibly gcc need manual intervention to
disable tmpfs but thats not often - if you dont mind the wasted disk
space (portage) you can avoid the sync time, but having the build
directories in tmpfs is a real gain - wish I had the max of 4G though :)
BillK
myth2 linux # df -h
Filesystem Size Used Avail Use% Mounted on
devtmpfs 878M 72K 878M 1% /dev
shm 878M 0 878M 0% /dev/shm
tmpfs 1.8G 492M 1.3G 29% /tmp
tmpfs 10M 0 10M 0% /var/lock
tmpfs 10M 64K 10M 1% /var/run
tmpfs 10M 320K 9.7M 4% /var/cache/hald
svcdir 2.0M 208K 1.8M 11% /var/lib/init.d
myth1:/home/MythTV/videos
1.2T 960G 251G 80% /mnt/videos
myth1:/home/MythTV/posters
1.2T 960G 251G 80% /mnt/posters
myth1:/home/MythTV/recordings
1.2T 960G 251G 80% /mnt/recordings
myth1:/home/MythTV/music
1.2T 960G 251G 80% /mnt/music
myth1:/home/MythTV/gallery
1.2T 960G 251G 80% /mnt/gallery
myth2 linux # swapon -s
Filename Type Size Used
Priority
/dev/nbd0 partition 2097148 0
-1
myth2 linux #
On Wed, 2010-12-15 at 10:15 +0100, Thomas Drueke wrote:
> Hi,
>
> is it possible to emerge packages to a $ROOT directory mounted via NFS ?
>
> The setup is
> - machine A is equipped with a Quad core CPU
> - machine B is equipped with an N330 Atom-CPU
> - machine A is doing the system update on a local chroot-environment
> for machine B and generates binary packages. These packages are
> installed on machine B using the binary package feature of portage.
>
> I expected that the above setup would give an performance improvement
> over letting machine B do the portage update itself. However a trial run
> did not show significant improvement that justifies the effort. Machine
> B still needs a reasonable amount of time to fetch unpack and install
> the packages.
>
> An alternative way might be to mount machine B's / directory via NFS
> and change make.conf's $ROOT variable to that mount point.
>
> Does that sound as a reasonable approach ?
>
> Regards,
> Thomas
>
--
William Kenworthy <[email protected]>
Home in Perth!
