ZFS in practice – verifying transparent compression –

I had read about the transparent compression that zfs offers. I wanted to verify it so I did, and sent my results off.

As we all know no experiment is valid unless verifiable. I wanted be sure that the transparent compression was actually working so I set up a sample zfs volume and purposefully made some files that i knew would compress really well. I grabbed their sizes, transferred them to another machine without zfs, and compared the size. It works.
filesystems usually suck

1. Create the zfs filesystem
root@machine1:~# zfs create rpool/ztest
root@machine1:~# zfs set mountpoint=/ztest rpool/ztest

as soon as you do this /ztest shows up in the filesystem
root@machine1:~# ls /ztest/

2. enable compression and set it use gzip at level 6 (default)
root@machine1:/ztest# zfs set compression=gzip rpool/ztest

3. create some files you know will compress well
time for i in $(seq 1 3); do echo -n “file${i} “; echo $(seq 1 10000000) >> file${i} 2>&1; done

4. check the compression ratio
root@machine1:/ztest# zfs get compressratio rpool/ztest
NAME         PROPERTY       VALUE        SOURCE
rpool/ztest  compressratio  3.69x        –

5. verify the compression is working by transferring the file to a machine without such an awesome FS, and checking the filesize there
root@machine1:/ztest# du -sh file1
23M     file1
scp -rvp /ztest/file1 root@machine2:/root/file1

root@machine2]# du -sh file1
83M    file1

Conclusion: transparent compression is working, and working rather well on a file I made to be easily compressed.

Found on Miro: Stump the Chef

I have been using Miro for a while now. Some of my favorite things to watch are things that would never make it to the ol’skool TV, but I think Stump the Chef has a shot. The theme of the show is as follows: The chef is presented with three seemingly random ingredients. He then has to take these three random ingredients, and mold something that the panel of three judges would pass as not only edible, but actually good. The difficulty of this task is usually exacerbated by how terribly the three ingredients would mesh traditionally. I really like the idea, its the kind of edutainment I wish was around when I was a TV watching kid. What do you think?

edit 01/2021 – miro is long gone but here are the stump the chef videos:

https://www.dailymotion.com/stumpthechef

Stateless Linux How To:

Stateless Linux sounds like a great idea for those of us that have many systems with similar configurations to manage. If you have never heard of the idea RTFM. I have stateless linux almost working here at my house. Here are the steps that I followed thus far.

on server system:
yum -y install cobbler
configure cobbler to manage dhcpd and start it
download Fedora 9 DVD ISO
sha1sum the iso to make sure the checksums match
mount the iso somewhere with the -o loop option eg. mount -o loop Fedora9.iso /media
import the distro into cobbler eg. cobbler import –mirror=/media/ –name=Fedora9
cp /sbin/lspci to /usr/sbin/lspci (if you skip this anaconda fails later under fedora 9)
mkdir -p /export/NFSroots/F9/
install os image into that directory eg. anaconda –headless -r /export/NFSroots/F9/ -m http://black/cobbler/ks_mirror/Fedora9/ –kickstart /var/lib/cobbler/kickstarts/f9base.ks
chroot into newly created environment
update environment (yum -y update)
create initrd
edit /etc/sysconfig/readonly-root
exit chroot
create cobbler distro
create cobbler profile
create cobbler system

on client system:
boot from lan.