I Do Linux

I'm Not Dead Yet

2013-11-13T10:23:00.001-05:00

I'm still here! But I've been busy with new responsibilities. Come, follow me: https://twitter.com/00gavin

Considering ZFS as a Home File Server

2013-03-30T17:03:00.000-04:00

I'm pondering building a home file server based on ZFS. Recently, many of the Linux distributions have started supporting ZFS, including Fedora and Scientific Linux.

If you're asking, "What the heck is ZFS?":

ZFS is a combined file system and logical volume manager designed by Sun Microsystems. The features of ZFS include protection against data corruption, support for high storage capacities, integration of the concepts of filesystem and volume management, snapshots and copy-on-write clones, continuous integrity checking and automatic repair, RAID-Z and native NFSv4 ACLs. --Wikipedia

Check out these links for the latest info on ZoL:

Note that this is not the slow FUSE userspace version of ZFS. This is a kernel module. I was considering installing the latest Fedora 18 to try this out, on some type of compact server. Ideally, a small, quiet, micro-ATX tower, with six 6Gbps SATA3 ports, would be best. Any recommendations? Maybe something like this:

StarCluster with Scientific Linux 6

2012-08-14T22:16:00.001-04:00

UPDATE: Please try the updated SL6 image ami-d60185bf to fix SSH key issues.

If you haven't heard of StarCluster from MIT, it is a toolkit for launching clusters of virtual compute nodes within the Amazon Elastic Compute Cloud (EC2). StarCluster provides a simple way to utilize the cloud for research, scientific, high-performance and high-throughput computing.

StarCluster defaults to using Ubuntu Linux (deb) images for its base, but I have prepared a Scientific Linux (rpm) image. Most of the research computing clusters I administrate are composed of Red Hat Enterprise, CentOS, or Scientific Linux, so I was eager to have a compatible cloud option built off of my currently favorite Linux flavor.

You can get started trying out StarCluster with the AWS Free Usage Tier. Once you have an Amazon account, StarCluster is easy to install with virtualenv and pip in your home directory.

$ mkdir starcluster ; cd starcluster
$ curl -O https://raw.github.com/pypa/virtualenv/master/virtualenv.py
$ python virtualenv.py foocluster
New python executable in foocluster/bin/python
Installing setuptools............................done.
Installing pip.....................done.
$ . foocluster/bin/activate
$ pip install starcluster
$ starcluster help

The first time starcluster is run, it will error without a config file, but prompt you to create one. Select option 2 to create and then edit the config file, setting preferred values. Make sure to substitute in your credentials. Here we are selecting micro instances of 64-bit Scientific Linux 6. Edit in and add the following lines to the ~/.starcluster/config file.

[aws info]
AWS_ACCESS_KEY_ID = your_aws_access_key_id
AWS_SECRET_ACCESS_KEY = your_secret_access_key
AWS_USER_ID= your_userid
[key foocluster]
KEY_LOCATION = ~/starcluster/foocluster.rsa
[cluster smallcluster]
KEYNAME = foocluster
NODE_IMAGE_ID = ami-d60185bf
NODE_INSTANCE_TYPE = t1.micro

Now we need to create the ssh key for access.

$ starcluster createkey -o foocluster.rsa foocluster

Now we can fire up a two-node cluster.

$ starcluster start -s 2 foocluster
$ starcluster listclusters
$ starcluster sshmaster foocluster
$ qstat -g c
$ exit
$ starcluster terminate foocluster

When done, make sure to go into the EC2 Management Console and terminate or delete any running instances or volumes. You don't want to run up a bill on idle instances.

Read more in the StarCluster manual or StarCluster wiki.

Here is the build log for my custom SL6 AMI. This image is based off of Jamie Kinney's Scientific Linux 6.2 x86_64 AMI.

$ starcluster start -s 1 foocluster -n ami-999d49f0
$ starcluster get foocluster /opt/sge6-fresh .
$ starcluster terminate foocluster
$ starcluster start -o -s 1 -i t1.micro -n ami-e2a0058b imagehost
$ starcluster listclusters --show-ssh-status imagehost
$ starcluster sshmaster imagehost -u ec2-user
$ sudo sed -i.bak -e's/\#PermitRootLogin\ yes/PermitRootLogin\ without-password/g' /etc/ssh/sshd_config
$ sudo sed -i.bak -e's/\#UseDNS\ yes/UseDNS\ no/g' /etc/ssh/sshd_config
$ sudo cp -f /home/ec2-user/.ssh/authorized_keys /root/.ssh/authorized_keys
$ sudo service sshd reload
$ exit
$ starcluster put imagehost ./sge6-fresh /opt/
$ rsync -e "ssh -i $HOME/starcluster/foocluster.rsa" -avP --delete sge6-fresh -l root ec2-23-23-64-53.compute-1.amazonaws.com:/opt/
$ starcluster sshmaster imagehost
# yum -y install yum-fastestmirror yum-conf-epel
# yum -y update
# yum -y install openmpi-devel nfs-utils-lib-devel java-1.6.0-openjdk-devel tmux zsh ksh tcsh \
  unzip mysql-server mysql httpd emacs ntsysv freetype-devel libpng-devel blas-devel lapack-devel \
  atlas-devel lbzip2 bzip2-devel ncurses-devel sqlite-devel zlib-devel libjpeg-devel mercurial ipython \
  python-imaging python-boto python-virtualenv Cython python-nose python-gnutls python-pip ruby
# echo "exit 0" >> /etc/init.d/portmap
# chmod +x /etc/init.d/portmap
# yum -y groupinstall "Development Tools"
# yum -y install infinipath-psm-devel
# yumdownloader --source openmpi
# rpm --import https://www.redhat.com/security/fd431d51.txt
# yum-builddep openmpi-1.5.4-1.el6.src.rpm
# rpm -ivh openmpi-1.5.4-1.el6.src.rpm
# rpmbuild -bb --define 'configure_options --with-sge' /root/rpmbuild/SPECS/openmpi.spec
# rpm -Uhv /root/rpmbuild/RPMS/x86_64/openmpi-1.5.4-1.el6.x86_64.rpm /root/rpmbuild/RPMS/x86_64/openmpi-devel-1.5.4-1.el6.x86_64.rpm
# echo "exclude=openmpi*" >> /etc/yum.conf
# . /etc/profile.d/modules.sh
# module load openmpi-x86_64
# ompi_info | grep -i grid
                 MCA ras: gridengine (MCA v2.0, API v2.0, Component v1.5.4)
# echo "module load openmpi-x86_64" >> /etc/profile.d/zzlocal.sh
# echo "module load openmpi-x86_64" >> /etc/profile.d/zzlocal.csh
# echo 'export PYTHONPATH=/usr/local/lib64/python2.6/site-packages:$PYTHONPATH' >> /etc/profile.d/zzlocal.sh
# echo 'setenv PYTHONPATH /usr/local/lib64/python2.6/site-packages:$PYTHONPATH' >> /etc/profile.d/zzlocal.csh
# chmod +x /etc/profile.d/zzlocal.*
# . /etc/profile.d/zzlocal.sh
# pip-python install --install-option="--prefix=/usr/local" mpi4py
# pip-python install --install-option="--prefix=/usr/local" numpy
# pip-python install --install-option="--prefix=/usr/local" scipy
# pip-python install --install-option="--prefix=/usr/local" matplotlib
# chmod u=rwx,go= /root /home/ec2-user
# vim /etc/rc.local # see below
# exit
$ starcluster listclusters
$ starcluster ebsimage i-xxxxxxxx starcluster-base-scientific-linux-6.2-x86_64-XX
>>> Your new AMI id is: ami-xxxxxxxx
$ starcluster terminate imagehost

/etc/rc.local

# update ec2-ami-tools
wget http://s3.amazonaws.com/ec2-downloads/ec2-ami-tools.noarch.rpm && \
rpm -Uvh ec2-ami-tools.noarch.rpm

# reset root password
dd if=/dev/urandom count=50|md5sum|passwd --stdin root
dd if=/dev/urandom count=50|md5sum|passwd --stdin ec2-user

# update root ssh keys
sleep 40
if [ ! -d /root/.ssh ]; then
    mkdir -p /root/.ssh
    chmod 700 /root/.ssh
fi
wget http://169.254.169.254/latest/meta-data/public-keys/0/openssh-key && \
cat openssh-key >>/root/.ssh/authorized_keys && \
chmod 600 /root/.ssh/authorized_keys
rm -f openssh-key

Scientific Linux 6.3 Released

2012-08-09T09:46:00.000-04:00

Scientific Linux version 6.3, the latest version of our favorite Red Hat Enterprise Linux compatible distribution, is out now. Get it while it's hot.

I'll be waiting for the release of the SL 6.3 live iso version, to boot off of a USB thumb drive with liveusb-creator. Currently, the SL 6.2 live iso and SL 5.7 live iso are still available for this purpose.

Fabcluster: An Amazon EC2 Script for HPC

2012-06-12T17:53:00.001-04:00

I've been experimenting with Amazon EC2 for prototyping HPC clusters. Spinning up a cluster of micro instances is very convenient for testing software and systems configurations. In order to facilitate training and discussion, my Python script is now available on GitHub. This piece of code utilizes the Fabric and Boto Python modules.

An Amazon Web Services account is required to use this script. Please note that running this script will start virtual servers in the cloud which will be billed for hourly. The AWS Free Usage Tier should be enough to get started, since the default instances are micro instances.

Let's install the fabcluster script, the required Python packages and grid engine files. Using pip and virtualenv is the easiest way to install Python packages as a non-root user.

$ git clone https://github.com/00gavin/fabcluster.git
$ cd fabcluster
$ curl -O https://raw.github.com/pypa/virtualenv/master/virtualenv.py
$ python virtualenv.py foo
$ . foo/bin/activate
$ pip install fabric
$ pip install boto

Once we have signed up for an Amazon Web Services account, place your credentials in the ~/.boto file:

[Credentials]
aws_access_key_id = YOUR Access Key ID HERE
aws_secret_access_key = YOUR Secret Access Key HERE

If we don't want to be prompted to accept ssh keys with every new instance, place these lines into the ~/.ssh/config file:

Host *amazonaws.com
        IdentityFile ~/.ssh/ec2-fab00-key.pem
        IdentityFile ~/.ssh/ec2-foo00-key.pem
        IdentityFile ~/.ssh/ec2-baz00-key.pem
        User root
        StrictHostKeyChecking no
        UserKnownHostsFile /dev/null

I like to use random variable names as prototype cluster names, like foo, baz, etc. Let's call our cluster "foo". If we have never ran this script before for a cluster named foo, we need to initialize a few security settings:

$ fab gridinit:foo

Now we can launch the head node of a cluster:

$ fab gridmake:foo

If we want to add two compute nodes to the foo cluster:

$ fab gridgrow:foo,2

To connect to our head node, we list our running instances and find the public hostname of foonode0.

$ fab gridlist
$ ssh ec2-NNN-NNN-NNN-NNN.compute-1.amazonaws.com
# qstat -g c
# exit

And when done, we want to make sure to terminate all running instances. Instances left running will be charged hourly.

$ fab gridterm:foo

A few things to note about this script:

uses the default Amazon Linux instance types which are Red Hat / Fedora based
defaults to the micro instance type but can be changed to the HPC cluster type
a normal user account is created matching your local username
uses Open Grid Scheduler
an example of initializing storage is commented out
parallel install of compute nodes is accomplished with subprocess

Update RHEL 5 with CentOS 5

2012-05-31T13:04:00.001-04:00

Now that Red Hat Enterprise Linux 6 is out, it may be desirable to update and convert non-critical Red Hat Enterprise Linux 5 servers to CentOS 5. Take note that the conversion process is to the same version. In this example, we see 64-bit RHEL 5:

# cat /etc/redhat-release 
Red Hat Enterprise Linux Server release 5.7 (Tikanga)
# uname -a
Linux foo.baz.edu 2.6.18-274.12.1.el5 #1 SMP Tue Nov 8 21:37:35 EST 2011 x86_64 x86_64 x86_64 GNU/Linux

# cd /usr/src
# wget http://mirror.centos.org/centos/5/os/x86_64/CentOS/yum-3.2.22-39.el5.centos.noarch.rpm
# wget http://mirror.centos.org/centos/5/os/x86_64/CentOS/yum-fastestmirror-1.1.16-21.el5.centos.noarch.rpm
# wget http://mirror.centos.org/centos/5/os/x86_64/CentOS/centos-release-5-8.el5.centos.x86_64.rpm
# wget http://mirror.centos.org/centos/5/os/x86_64/CentOS/centos-release-notes-5.8-0.x86_64.rpm
# cp /etc/redhat-release /etc/redhat-release.redhat
# yum -y remove rhn*
# rpm --import http://mirror.centos.org/centos/RPM-GPG-KEY-CentOS-5
# rpm -Uhv yum-3.2.22-39.el5.centos.noarch.rpm yum-fastestmirror-1.1.16-21.el5.centos.noarch.rpm
# rpm -ev --nodeps redhat-release
# rpm -ihv centos-release-5-8.el5.centos.x86_64.rpm centos-release-notes-5.8-0.x86_64.rpm
# yum update
# vim /etc/grub.conf
# reboot

Building ScaLAPACK with LAPACK, BLAS and Open MPI

2012-03-30T10:44:00.001-04:00

ScaLAPACK is a library of high-performance linear algebra routines for parallel distributed memory machines. ScaLAPACK solves dense and banded linear systems, least squares problems, eigenvalue problems, and singular value problems.

This build log uses the Intel compiler suite, which is free to download for non-comercial use. I recommend installing Intel Parallel Studio XE for Linux. This build is on a 64-bit x86_64 Scientific Linux 6 system, but these instructions should also work on Red Hat Enterprise Linux and CentOS 5 & 6, too.


# cd /usr/local/
# wget http://www.netlib.org/blas/blas.tgz
# tar xzvf blas.tgz
# mv BLAS blas-20110419
# ln -s blas-20110419 BLAS
# cd BLAS
# ifort -FI -w90 -w95 -cm -O3 -unroll -c *.f
# ar r libfblas.a *.o
# ranlib libfblas.a
# rm -rf *.o
# ln -s libfblas.a libblas.a
# export BLAS=/usr/local/BLAS/libblas.a


# cd /usr/local/
# wget wget http://www.netlib.org/lapack/lapack.tgz
# tar xzvf lapack.tgz
# ln -s lapack-3.4.0 LAPACK
# cd LAPACK
# cp INSTALL/make.inc.ifort make.inc
# make lapacklib
# make clean
# export LAPACK=/usr/local/LAPACK/liblapack.a


# cd /usr/local/src
# wget http://www.open-mpi.org/software/ompi/v1.4/downloads/openmpi-1.4.5.tar.gz
# tar xzvf openmpi-1.4.5.tar.gz
# cd openmpi-1.4.5
# CC=icc FC=ifort F77=ifort CXX=icpc ./configure --prefix=/usr/local/openmpi-1.4.5
# make
# make install
# cd /usr/local
# ln -s openmpi-1.4.5 openmpi
# export PATH=/usr/local/openmpi/bin:$PATH
# export LD_LIBRARY_PATH=/usr/local/openmpi/lib:$LD_LIBRARY_PATH


# cd /usr/local
# wget http://www.netlib.org/scalapack/scalapack-2.0.1.tgz
# tar xvf scalapack-2.0.1.tgz
# ln -s scalapack-2.0.1 scalapack
# cd scalapack
# cp SLmake.inc.example SLmake.inc
# vim SLmake.inc
    BLASLIB       = -lblas -L/usr/local/BLAS
    LAPACKLIB     = -llapack -L/usr/local/LAPACK
    LIBS          = $(LAPACKLIB) $(BLASLIB)
# make
# export SCALAPACK=/usr/local/scalapack/libscalapack.a

References:

Fedora 16 Xfce on an Eee PC

2012-01-19T22:04:00.002-05:00

Here are some notes from a recent install of Fedora 16, 64-bit, on an Asus Eee PC 1015PEM Seashell netbook, with a 2GB memory upgrade and an SSD. This system has a 64-bit 1.5 GHz Intel Atom N550 processor. I skipped Gnome 3 and went with the Fedora Xfce Spin.

Create installation media by downloading the ISO of the Fedora 16 Xfce Spin 64-bit. A bootable USB drive can be prepared with liveusb-creator. To boot the installer, insert the prepared liveusb drive into the left USB port and hit the Esc key after powering on. This brings up the boot menu where the USB drive can be selected. The Linux install should be mostly defaults, except that disk encryption and a grub password should be enabled.

After a fresh install and reboot, plug into a wired network connection, install the RPM Fusion software repository, do the first big system software update, then install the wireless driver package. Note that kmod-wl from RPM Fusion provides a driver for the Broadcom wireless card after reboot.

# yum -y install yum-fastestmirror
# yum -y update
# yum localinstall --nogpgcheck \
http://download1.rpmfusion.org/free/fedora/rpmfusion-free-release-stable.noarch.rpm \
http://download1.rpmfusion.org/nonfree/fedora/rpmfusion-nonfree-release-stable.noarch.rpm
# yum install kmod-wl akmod-wl

and install some extra packages you may or may not find useful...

# yum -y install krb5-workstation tkinter mutt getmail \
python-setuptools ipython mysql-workbench gconf-editor git \
subversion ethtool liveusb-creator telnet freeglut-devel \
gcc gcc-c++ SDL-devel rdesktop thunderbird thunderbird-enigmail \
vim fortune-mod lsb redhat-lsb-graphics qt-x11 powertop dosbox \
tigervnc tigervnc-server audacity-freeworld rdiff-backup \
samba-client rtorrent unrar anki libreoffice-langpack-de \
libreoffice-writer libreoffice-calc libreoffice-impress \
screen nmap net-snmp-utils wireshark-gnome iptraf strace sysstat \
gimp rpmdevtools yum-utils ncurses-devel dvd+rw-tools \
libpng-devel mednafen jwhois java-1.6.0-openjdk-plugin wget \
ddrescue wodim vlc mplayer mencoder grip lame easytag \
lm_sensors xterm xorg-x11-xinit-session xorg-x11-apps

The biggest tip with a 10" netbook is to consolidate and auto-hide menus/toolbars for maximum screen real estate. And when sitting at a desk, use a full-sized LCD monitor. The screen space issue is one of the main reasons I went with Xfce over Gnome 3. Gnome 3 wastes too much screen space with wide toolbars and window borders.

The trackpad is nice, supporting one, two and three-finger tapping. Enable this in the Xfce mouse preferences is not possible, though. Adding three lines to a config file fixes this, for easy copy-and-paste the way it was meant to be. On Linux, mouse button-one is highlight/select, copy being automatic, and button-three is paste. With a touchpad this is a double-tap with one finger and drag to select text, then a two-finger tap to paste.

/usr/share/X11/xorg.conf.d/50-synaptics.conf

Section "InputClass"
        Identifier "touchpad catchall"
        Driver "synaptics"
        MatchIsTouchpad "on"
        MatchDevicePath "/dev/input/event*"
        Option "TapButton1" "1"
        Option "TapButton2" "2"
        Option "TapButton3" "3"
EndSection

Now, one could stop here and be finished, but I'm interested in tweaking for power saving features, like the Super Hybrid Engine (SuperHE) features of this model. To enable the Asus SuperHE power saving, we add a few scripts to the power management system.

/etc/pm/power.d/asus_she_power

#!/bin/sh
# /etc/pm/power.d/asus_she_power

cpufv="/sys/devices/platform/eeepc-wmi/cpufv"
lcd="/sys/devices/virtual/thermal/cooling_device4/cur_state"

case "$1" in
    true)
        echo "Super Hybrid Engine ENABLED"
        echo 2 > $cpufv
        echo 8 > $lcd
    ;;
    false)
        echo "Super Hybrid Engine DISABLED"
        echo 0 > $cpufv
        echo 0 > $lcd
        setpci -s 00:02.0 f4.b=ff
    ;;
esac

exit 0

/etc/pm/sleep.d/asus_she_sleep

#!/bin/sh
# /etc/pm/sleep.d/asus_she_sleep
 
battery="/proc/acpi/ac_adapter/AC0/state"
ac="/sys/bus/acpi/drivers/ac/ACPI0003:00/power_supply/AC0/online"
 
case "$1" in
    thaw)
        if (test -e $battery); then
            if grep off-line "$battery" > /dev/null; then
                /etc/pm/power.d/asus_she_power true
            fi
        elif (test -e $ac); then
            if grep 0 "$ac" > /dev/null; then
                /etc/pm/power.d/asus_she_power true
            fi
        else
            /etc/pm/power.d/asus_she_power false
        fi
esac
 
exit 0

These two above scripts will trigger when plugging power and waking from suspend respectively. Don't forget to make them executable:

# chmod +x /etc/pm/sleep.d/asus_she_sleep /etc/pm/power.d/asus_she_power

With an SSD, we want to enable the noop scheduler. Edit /etc/default/grub and add the elevator option to the kernel line:

GRUB_TIMEOUT=5
GRUB_DISTRIBUTOR="Fedora"
GRUB_DEFAULT=saved
GRUB_CMDLINE_LINUX="rd.md=0 rd.dm=0  KEYTABLE=us rd.lvm.lv=vg_foo/lv_swap quiet rd.luks.uuid=luks-xxxxxxxx-xxxx-xxxx-xxxx SYSFONT=latarcyrheb-sun16 rhgb rd.lvm.lv=vg_foo/lv_root LANG=en_US.UTF-8 elevator=noop"

And rebuild the grub configuration:

# grub2-mkconfig -o /boot/grub2/grub.cfg

Enable noatime on ext4 file systems, and tmpfs for temporary directories in the /etc/fstab file:

/dev/mapper/vg_foo-lv_root / ext4 defaults,noatime 1 1
tmpfs /tmp tmpfs defaults,nodev,nosuid,size=256M 0 0
tmpfs /var/tmp tmpfs defaults,nodev,nosuid,size=64M 0 0
tmpfs /var/log tmpfs defaults,nodev,nosuid,size=32M 0 0

Some additional power and SDD saving tweaks:

# touch /etc/rc.d/rc.local
# chmod +x /etc/rc.d/rc.local

/etc/rc.d/rc.local

echo 1 > /proc/sys/vm/swappiness
echo 50 > /proc/sys/vm/vfs_cache_pressure
echo 1500 > /proc/sys/vm/dirty_writeback_centisecs
echo 20 > /proc/sys/vm/dirty_ratio
echo 10 > /proc/sys/vm/dirty_background_ratio
echo 0 > /proc/sys/kernel/nmi_watchdog
echo 1 > /sys/devices/system/cpu/sched_smt_power_savings
echo 1 > /sys/module/snd_hda_intel/parameters/power_save
echo 5 > /proc/sys/vm/laptop_mode
#echo noop > /sys/block/sda/queue/scheduler 
#for I in /sys/devices/system/cpu/cpu?/cpufreq/scaling_governor; do echo 'ondemand' >$I; done
#echo min_power > /sys/class/scsi_host/host0/link_power_management_policy
#ethtool -s eth0 wol d
#modprobe sdhci
#for I in `find /sys -name autosuspend -exec echo {} \;`; do echo "0" >"$I"; done
for I in `find /sys/devices -wholename "*pci*power/control" |grep -v "usb2"`; do echo 'auto' >"$I"; done

Disable some unused services on a laptop client machine:

# for s in abrtd-ccpp abrt-oops abrt-vmcore abrtd atd auditd avahi-daemon fcoe ip6tables iscsi livesys-late livesys lldpad mdmonitor-takeover sendmail sm-client; do echo $s; systemctl disable $s.service; done

If restoring a home directory from backup, with SELinux enabled, make sure to restore the security context of the home directories:

# restorecon -vr /home

Google Chrome is an excellent browser for a netbook. It is a very efficient in terms of performance speed and screen utilization. Download the rpm package from the Google website, then install it with yum to get all dependencies.

$ sudo yum install ~/Downloads/google-chrome-stable_current_x86_64.rpm

Skype is good for video chat and international voice calling. Download the RPM from the Skype website, then use yum to install.

$ sudo yum install ~/Downloads/skype-2.2.0.35-fedora.i586.rpm

I am a big fan of video calling on Linux netbooks and Android phones. If you use Google Talk for chat, the voice and video plugin will allow easy video chat and conferencing with Google+ Hangouts.

$ sudo yum install ~/Downloads/google-talkplugin_current_x86_64.rpm

To get the latest version of 64-bit Flash, install the Adobe repo, then the plugin package.

# yum install http://linuxdownload.adobe.com/adobe-release/adobe-release-x86_64-1.0-1.noarch.rpm
# yum install flash-plugin

Install some i686 libraries for 32-bit compatibility.

# yum install libXmu.i686 libXt.i686

If using Thunderbird for mail, you may have to tell it to default to Chrome for web links with this quick fix:

$ xdg-mime query default x-scheme-handler/http
$ xdg-mime default google-chrome.desktop x-scheme-handler/http
$ xdg-mime default google-chrome.desktop x-scheme-handler/https

You can change video modes from the command line when connecting to an external monitor or video projector. Stick these aliases into your ~/.bashrc file:

alias vga='xrandr --fb 1280x1624 --output LVDS1 --primary --mode 1024x600 --rate 60 --pos 0x0 --panning 0x0 --output VGA1 --mode 1280x1024 --rate 60 --pos 0x1024 --above LVDS1'
alias mir='xrandr --fb 1280x1024 --output LVDS1 --primary --mode 1024x600 --rate 60 --pos 0x0 --panning 1280x1024 --output VGA1 --mode 1280x1024 --rate 60 --pos 0x0 --same-as LVDS1'
alias lcd='xrandr --fb 1024x600  --output LVDS1 --primary --mode 1024x600 --rate 60 --output VGA1 --off'

After using Xfce for a few days, I'm liking it.

For further details on Fedora 16, I recommend reading the Release Notes, particularly the section on Changes for System Administrators.

Grid Engine Config Tips

2011-12-17T16:30:00.000-05:00

Here are some of the Grid Engine configuration steps we should take on a new install. I recommend doing all of these from the very beginning, to prevent changes that may confuse or break user workflow.

There is one thing we must always do with a new compute cluster, and that is enable hard memory limits. Users are usually not too keen on any kind of limit, because jobs will eventually run into them. Once the realization is made that limits ensures node stability and uptime, users will demand them. Without limits, one bad job can crash a node and bring down many other jobs.

To enable hard memory limits, we modify the complex configuration to make h_vmem requestable.

# qconf -mc
h_vmem              h_vmem     MEMORY      <=    YES         YES        1g       0

Once this complex is set, it is a good idea to define a default option for qsub in the $SGE_ROOT/default/common/sge_request file. When enabling h_vmem, we should also set a default value for h_stack. h_vmem sets a limit on virtual memory, while h_stack sets a limit on stack space for binary execution. Without a sufficient value for h_stack, programs like Python, Matlab or IDL will fail to start. Here, we are also binding each job to a single core.

-binding linear:1
-q all.q
-l h_vmem=1g
-l h_stack=128m

If we want to manually set values for each individual node, like slots and memory, a for-loop is very helpful.

# qconf -rattr exechost complex_values slots=8,num_proc=8,h_vmem=8g node01
# for ((I=1; I <= 16 ; I++)); do 
> NODE=`printf "node%02d\n" $I`
> MEM=`ssh $NODE 'free -b |grep Mem |cut -d" " -f 5'`
> SWAP=`ssh $NODE 'free -b |grep Swap |cut -d" " -f 4'`
> VMEM=`echo $MEM+$SWAP|bc`
> qconf -rattr exechost complex_values slots=8,num_proc=8,h_vmem=$VMEM $NODE
> done

To submit a job with a 4 gig limit, use the -l command line option.

$ qsub -l h_vmem=4g -l h_stack=256m myjob.sh

To see available memory, use qstat.

$ qstat -F h_vmem

It is also a good idea to place limits on the amount of memory any single process on the login node may allocate, in the /etc/security/limits.conf file. This example will limit any user in the clusterusers group to 4 gigs per process. Anything larger should be ran via qlogin. When adding new users, make sure to add them to this now default group.

# limit any process to 4GB = 1024*1024*4KB = 4194304
@clusterusers      hard    rss             4194304
@clusterusers      hard    as              4194304

There should also be a limit on how many jobs a single user can queue at once. If a user must submit over 2000 jobs simultaneously, they may want to consider a more manageable workflow utilizing array jobs.

# qconf -mconf
max_u_jobs 2000

If we want to limit the number of jobs a single user can have in the running state simultaneously:

# qconf -msconf
 max_reservation 128
 maxujobs 128

If the queue will be accepting multi-slot parallel jobs, slot reservation should be enabled to prevent starvation. Otherwise, single-slot jobs will constantly fill in space ahead of the big job. This can be done by submitting multi-slot jobs with the “-R y" option.

To enable a simple fairshare policy between all users, there are only three options to check:

# qconf -mconf
enforce_user auto
auto_user_fshare 100
# qconf -msconf
weight_tickets_functional 10000

To be a bit more verbose, we should collect some job scheduler info.

# man sched_conf
# qconf -msconf
 schedd_job_info true

Now we can see why or why not a job is scheduled.

$ qstat -j 427997
$ qacct -j 427997

If we plan to allow graphical GUI programs in the queue, we must setup a qlogin wrapper script with proper X11 forwarding.

# vim /usr/global/sge/qlogin_wrapper
# chmod +x /usr/global/sge/qlogin_wrapper

qlogin_wrapper:

#!/bin/sh
HOST=$1
PORT=$2
shift
shift
echo /usr/bin/ssh -Y -p $PORT $HOST
/usr/bin/ssh -Y -p $PORT $HOST

Set the qlogin wrapper and ssh shell:

# qconf -mconf
 qlogin_command /usr/global/sge/qlogin_wrapper
 qlogin_daemon /usr/sbin/sshd -i

If we have a floating license server with a limited number of seats, we will want to configure a consumable complex resource. When a user submits a job, the qsub option '-l idl=1' must be used. In this example, the number of jobs that specify idl will be limited to 15 at any one time.

# qconf -mc
matlab ml INT <= YES YES 0 0
idl idl INT <= YES YES 0 0
# qconf -me global
complex_values matlab=10,idl=15

If we want to have multiple queues across the same hosts, we can define a policy so that nodes do not become oversubscribed.

# qconf -arqs
{
    name         limit_slots_to_cores_rqs
    description  Prevents core oversubscription across queues. 
    enabled      TRUE    
    limit        hosts {*} to slots=$num_proc
}

Here are some handy aliases I find useful in my ~/.bashrc file:

alias qconf-all='qconf -mq all.q'
alias qconf-offline='echo "all.q@nodeXX" ; \qconf -rattr queue slots 0' # all.q@node23
alias qconf-online='echo "all.q@nodeXX" ; \qconf -rattr queue slots 8' # all.q@node23
alias qstat-errors='\qstat -f -explain E'
alias qstat-summary='\qstat -g c'
alias qstat-mem='qstat -F h_vmem'
alias qstat-ext='qstat -ext'
alias qstat-io='qstat -ext | awk '"'"'{print $11 "  " $5 "  " $1}'"'"' | grep -v "\-\-" | sort -n'
alias qmod-clear='\qmod -c "*"'

And done. If you have some must-do configuration steps, please post them in the comments bellow.

GPGPU with AMD

2011-12-14T19:41:00.003-05:00

GPGPU (General Purpose Graphics Processing Unit) is a high performance computing solution which off-loads heavy floating-point computation to the video card. OpenCL is a cross-platform cross-vendor solution for parallel computing with various kinds of coprocessors such a GPU video card.

Here, we will be configuring an AMD FireStream 9250 GPU Compute Accelerator which provides over 1 TFLOPS of computing power. The target operating system is Scientific Linux 6 x86_64, which is Red Hat Enterprise Linux 6 compatible.

You can verify the model of GPU installed with the lspci command:

# lspci | grep VGA
07:00.0 VGA compatible controller: ATI Technologies Inc RV770 [FireStream 9250]

To find the latest graphics driver, head over to the AMD driver download page. The current version is the AMD Catalyst 11.12 Proprietary Linux x86 Display Driver. Once we have the URL of the driver we can begin.

# mkdir /usr/local/src/amd ; cd /usr/local/src/amd
# wget http://www2.ati.com/drivers/linux/ati-driver-installer-11-12-x86.x86_64.run
# chmod +x ati-driver-installer-11-12-x86.x86_64.run
# ./ati-driver-installer-11-12-x86.x86_64.run --buildpkg RedHat/RHEL6_64a
# yum -y install fglrx64_p_i_c-8.92-1.x86_64.rpm
# aticonfig --initial -f

We may need to tweak the GDM setting to allow remote users access to the GPU with the /etc/gdm/custom.conf config file:

[servers]
0=Rendering
[server-Rendering]
command=/usr/bin/Xorg -br -ac -audit 0
flexible=true

An we also may need to allow permissions to direct rendering for shell users in the /etc/X11/xorg.conf config file:

Section "DRI"
Mode 0666
EndSection

Now to restart the display and verify that the FireGL module is in use.

# killall -9 gdm-binary
# lsmod | grep fgl
fglrx                3154329  49
# grep FireGL /var/log/Xorg.0.log
(II) Module fglrx: vendor="FireGL - ATI Technologies Inc."
(II) Module fglrxdrm: vendor="FireGL - ATI Technologies Inc."
(II) fglrx(0):     Desc: ATI FireGL DRM kernel module

Install the AMD APP SDK v2.5 with OpenCL:

# wget http://developer.amd.com/Downloads/AMD-APP-SDK-v2.5-lnx64.tgz
# mkdir AMD-APP-SDK-v2.5-lnx64 ; cd AMD-APP-SDK-v2.5-lnx64
# tar xzvf 
# ./Install-AMD-APP.sh

Install GPU-enabled BLAS and FFT, the AMD Accelerated Parallel Processing Math Libraries (APPML) v1.4.

# cd /usr/local/src/amd
# wget http://developer.amd.com/Downloads/clAmdBlas-1.4.182-Linux.tar.gz
# wget http://developer.amd.com/Downloads/clAmdFft-1.4.182-Linux.tar.gz
# wget http://developer.amd.com/Downloads/LUDOpenCLBLAS-Linux.zip
# wget http://developer.amd.com/Downloads/ObjectDetection-Linux.zip
# mkdir clAmdBlas-1.4.182-Linux ; cd clAmdBlas-1.4.182-Linux
# tar xzvf ../clAmdBlas-1.4.182-Linux.tar.gz
# ./install-clAmdBlas-1.4.182.sh
# cd ..
# mkdir clAmdFft-1.4.182-Linux ; cd clAmdFft-1.4.182-Linux
# tar xzvf ../clAmdFft-1.4.182-Linux.tar.gz
# ./install-clAmdFft-1.4.182.sh

Verify everything is functioning.

# . /etc/profile.d/ati-fglrx.sh
# . /opt/clAmdBlas-1.4.182/appmlEnv.sh
# /opt/clAmdBlas-1.4.182/bin64/clBlasVersion 
clAmdBlas version 1.4.182
# /opt/clAmdBlas-1.4.182/bin64/example_sgemm 
Result:
21370 22040 22710 
37070 38240 39410 
52770 54440 56110 
68470 70640 72810 
# /opt/clAmdBlas-1.4.182/bin64/example_strsm 
Result:
-1.17478e+00 -7.83188e-01 -3.91593e-01 1.51806e-06 3.91595e-01 
-4.03831e-01 -2.69220e-01 -1.34610e-01 5.98375e-07 1.34612e-01 
-2.06611e-01 -1.37741e-01 -6.88705e-02 2.98023e-07 6.88705e-02 
9.31818e+00 9.54545e+00 9.77273e+00 1.00000e+01 1.02273e+01
# . /opt/clAmdFft-1.4.182/appmlEnv.sh
# /opt/clAmdFft-1.4.182/bin64/clAmdFft.Client
                Client Test *****PASS*****

Read more about it all at the AMD OpenCL Zone.

How Much Work Can a Cluster Do?

2011-11-01T16:00:00.000-04:00

If we take a 32 node computer cluster, with each compute node averaging over 2 amps under high load, add in some extra amps for storage and network components, all running on 208 volt, we could be pulling over 90 amps total.

current x voltage = power
 amps  x  volts   = watts
   90A  x  208V   = 18720W

Wikipedia say, "A laborer over the course of an 8-hour day can sustain an average output of about 75 watts." So a 32 node HPC cluster is roughly equivalent to 250 laborers that never sleep!

Watt - Wikipedia

Easy Backup of MySQL Databases

2011-10-12T15:03:00.000-04:00

Real Men don't make backups. They upload it via ftp and let the world mirror it.
-- Linus Torvalds

Here is a simple method for doing daily backups of all MySQL databases on a server. This method will save a separate file for each database, every morning at 4AM, into a file named with the host, database and date. This method is extremely convenient for quick restores, to any day within the last two weeks.

Create a mysql backup user and grant database permissions. Then set the backup script to run every day at 4AM with cron.

# adduser mysqlbackup
# mysql -u root -p -e "GRANT SELECT, LOCK TABLES ON *.* TO 'mysqlbackup'@'localhost' IDENTIFIED BY 'XXXXXXX'; flush privileges;"
# touch /home/mysqlbackup/mysqlbackup.sh ; chmod +x /home/mysqlbackup/mysqlbackup.sh
# echo "0 4 * * * /home/mysqlbackup/mysqlbackup.sh" |crontab -u mysqlbackup -

Set the backup user mysql password, replacing XXXXXX with your own password, in the .my.cnf file.

/home/mysqlbackup/.my.cnf

[client]
user=mysqlbackup
password=XXXXXXX

/home/mysqlbackup/mysqlbackup.sh

#!/bin/sh

### get the hostname
MYHOST="`hostname -s`"

### individually dump all databases with timestamps
for I in `mysql -e "show databases;" |egrep -v '^Database$|hold$'`; do
        ### dump and compress
        mysqldump --single-transaction --quote-names --opt $I |gzip -cf >/home/mysqlbackup/$MYHOST-$I-`date +%Y%m%d`.sql.gz
        ### remove two-week old dumps
        rm -f /home/mysqlbackup/$MYHOST-$I-`date +%Y%m%d --date="-14 days"`.sql.gz
done

I recommend running the script at least once by hand to check functionality.


# su - mysqlbackup
$ /home/mysqlbackup/mysqlbackup.sh
$ exit

Now that you have daily database dumps, to do a restore you simply cat the dump file to the mysql client. The current database will be dropped and rebuilt from the file selected. Here is an example of restoring the database importantstuff on host foo from 4AM on 2011, April 30th.

# zcat /home/mysqlbackup/foo-importantstuff-20110430.sql.gz |mysql -u root -p importantstuff

It's easy. Do your backups!

Auto Update Kernel Modules with DKMS

2011-10-05T15:49:00.000-04:00

I had a brand new Dell OptiPlex 990 desktop computer that had just been installed with Scientific Linux 6. Enterprise Linux runs great on these things, except I was getting a lot of network errors.

# lspci | grep Ethernet
00:19.0 Ethernet controller: Intel Corporation 82579LM Gigabit Network Connection (rev 04)

Sep 11 03:59:47 foo kernel: e1000e: em1 NIC Link is Down
Sep 11 03:59:47 foo NetworkManager[1760]:  (em1): carrier now OFF (device state 8, deferring action for 4 seconds)
Sep 11 03:59:47 foo kernel: e1000e 0000:00:19.0: em1: Reset adapter

The solution was to update to the latest Intel Wired Ethernet e1000e driver, currently version 1.6.2. It is easy enough to download, un-tar, and make install the package, but it will be ignored after a new kernel is installed.

Dynamic Kernel Module Support (DKMS) is an easy solution to such a problem. Since we are replacing a kernel module that is included in the stock kernel RPM package, we want to be sure that the updated driver will be built and installed automatically whenever a new kernel is booted.

DKMS is easy to install with the EPEL repository enabled. Here we are enabling EPEL on Scientific Linux 6, then yum installing the DKMS package:

# rpm -ihv http://download.fedoraproject.org/pub/epel/6/x86_64/epel-release-6-5.noarch.rpm
# yum install dkms

We must download and prepare the latest module source code by moving the source code to a /usr/src/MODULENAME-VERSION directory, then create a dkms.conf file within:

# cd /usr/local/src
# wget http://downloads.sourceforge.net/project/e1000/e1000e%20stable/1.6.2/e1000e-1.6.2.tar.gz
# tar xzvf e1000e-1.6.2.tar.gz
# mv e1000e-1.6.2/src /usr/src/e1000e-1.6.2
# vim /usr/src/e1000e-1.6.2/dkms.conf

dkms.conf:

PACKAGE_NAME="e1000e"
PACKAGE_VERSION="1.6.2"
BUILT_MODULE_NAME[0]="e1000e"
DEST_MODULE_LOCATION[0]="/kernel/drivers/net/e1000e/"
AUTOINSTALL="yes"

Now that we have the source code in place, we tell DKMS to maintain, build and install the module:

# dkms add -m e1000e -v 1.6.2
# dkms build -m e1000e -v 1.6.2
# dkms install -m e1000e -v 1.6.2
# chkconfig dkms_autoinstaller on

We must change the search order for modules to allow "weak modules" to load before the built-in modules by editing the /etc/depmod.d/dist.conf file:

search updates extra weak-updates built-in

This last step ensures that when rebooting into a new kernel, the custom module will be loaded instead of the RPM packaged module.

NX - A Fast Remote Linux Desktop

2011-08-30T14:59:00.018-04:00

Many Linux users are familiar with X11 forwarding via SSH, or VNC for remote desktop sharing. SSH can tunnel X Windows from your remote host to the local X server display. And VNC can remotely share the desktop seen on the monitor or start a completely new desktop in the background. Similarly, NoMachine (NX) can do the same things, with the bonus of a faster compression algorithm for a more responsive GUI.

I had avoided NX because it seemed confusing and difficult to install. Now, NX has a free client, node and server download, all of which is easy to install and use. These steps work on the latest CentOS 5 and Fedora 15.

Download the latest NX Free Edition client, node and server RPMs from www.nomachine.com/download.php and install them on your server:

# rpm -ihv nxclient-3.5.0-7.x86_64.rpm nxnode-3.5.0-6.x86_64.rpm nxserver-3.5.0-8.x86_64.rpm

On your client system, just install the client RPM:

# rpm -ihv nxclient-3.5.0-7.x86_64.rpm

Now you may log into your server by providing your username and password.

NX provides a much more responsive remote desktop, especially over low speed and high latency connections. There are also many options for choosing Gnome vs KDE, or even just a single program for your session. Give it a try.

There is also an option for sharing your desktop with multiple other users, called shadowing in NX terminology. The primary user logs in and starts a new Unix session. Then another user logs in with a Shadow session, selecting the session of the primary user. This is great for remote collaboration with trusted colleagues that have an account on the same server.

Make sure to check out Getting Started with NX. Also, detailed instructions and troubleshooting for installing the repository version of NX can be found at CentOS Wiki HowTos/FreeNX.

UPDATE: I have found that the free NX server download from nomachine.com is more functional than the one provided in the repositories. In particular, shadowing was broken in the repo version. Shadowing is the desktop session sharing feature of NX. So, I do not recommend using yum to install NX. NX Free Edition seems to be limited to two users, though.

LinuxCon Live Stream

2011-08-17T15:46:00.001-04:00

Didn't get a plane ticket to Vancouver for LinuxCon North America 2011? Well, there is a live video stream.

North America 2011 | Live Video Streaming

CentOS 6 Available

2011-07-11T09:39:00.005-04:00

One of my favorite, license fee-free, Red Hat Enterprise Linux rebuilds is now available. CentOS is completely RHEL-compatible, with a different logo. Read all about it:

Of particular note is that the i386 DVD image is a bit too big to fit on a DVD+R, so use a DVD-R. When the live CDs are released in a few days, be sure to give liveusb-creator a try to make a bootable USB thumb drive.

If you want a Linux distribution with great backwards compatibility, solid updates, a long support cycle (7+ years), that is tested on enterprise server hardware, then CentOS should be at the top of your list. It turns out that CentOS is the distribution of choice for 30% of web servers.

Fedora 15 is Out

2011-05-25T11:05:00.005-04:00

The latest and greatest of Linux, Fedora 15, has been released. Get your copy now!

Check out the release notes for more info.

So I tried Fedora 15 beta with the new Gnome 3 desktop, on my netbook. I didn't like the default Gnome desktop. I reverted back to Fedora 14. A netbook really needs a screen-space efficient window manager. I want all toolbars to be consolidated and hide when not needed. So what are other's doing? Time to switch to the LXDE or XFCE spins?

Video Tour of the Facebook Server Room

2011-04-21T11:38:00.003-04:00

If you didn't already know, Facebook runs on a Linux software stack which utilizes Apache, MySQL, Memcached, PHP, as well as many other open source technologies.

Facebook also happens to be doing some interesting things in their data centers, with the Open Data Center Alliance and Open Compute Project.

See behind the scenes in this video of Facebook data center hot isle containment.

via Data Center Knowledge

20 Year of Linux

2011-04-12T12:59:00.001-04:00

Grid Engine is Alive and Well

2011-04-08T13:54:00.002-04:00

Now that the pieces are falling into place, after Oracle's acquisition of Sun Microsystems, we can see that open source can vendor-proof central pieces of your software infrastructure. Grid Engine is alive and well, with a blog, a mailing list, a git repository, and even commercial support from Univa.

I'd like to give a big THANK YOU to Dag over at Bioteam, for being so active and providing courtesy binaries for CentOS Linux and Mac OS X.

Source code repositories can also be found from the Open Grid Scheduler and Son of Grid Engine projects.

Speech Input

2011-03-23T15:17:00.005-04:00

Testing speech input with Chrome 11. Neat.

Amazon EC2 for HPC

2011-03-15T12:32:00.007-04:00

I've been sitting on this for a while now. I did quite a lot of testing with Amazon EC2, Elastic Cloud Computing, to determine if replacing an on-site rack of high performance computing cluster hardware with cloud computing resources is feasible. Utilizing the Fabric and Boto Python modules, I have build a command line utility for automating cluster deployments within the cloud. The cluster utilizes a Red Hat Enterprise Linux-compatible operating system with Grid Engine for job queuing. Initial testing targeted the large HPC instance types, but I have since been able to substitute in micro instances for faster and more economical testing of deployments.

This slideshow covers the the basic terminology, and compares the cost of purchasing and maintaining a rack of equipment versus cloud fees.

Executive summary: Yes, an on-site rack of equipment can be replaced with cloud computing resources, while continuing to utilize existing workflows. The cost to a university for EC2 is competitive with in-house hardware, when factoring in data center and maintenance costs. The cost to faculty is prohibitively expensive for cloud, since university overhead is paid up-front. Data locality is a big concern. An added benefit of cloud computing is hardware abstraction, meaning that staff time is focused instead on system configuration and software deployment.

Fedora 14 on an Eee PC 1015PEM

2011-03-09T16:26:00.041-05:00

Here are some notes from a recent install of Fedora 14, 32-bit, on an Asus Eee PC 1015PEM Seashell netbook. This system has a 64-bit 1.5 GHz Intel Atom N550 processor, but things like Skype may have problems with 64-bit. I'll go 64-bit when Fedora 15 is released.

The 1GB memory chip was upgraded to a Crucial 2GB, 204-pin SODIMM, DDR3 PC3-10600 memory module. This was easily done by unscrewing the small panel on the bottom of the system and un-clipping the existing chip. The hard disk drive was also upgraded to a Corsair Nova 32GB 2.5" SSD, which was not-so-easily done. Replacing the drive involving removing the keyboard, prying apart plastic and removing many screws. I thought I was going to snap it to pieces! A tool, like the one in these disassembly photos, would have been handy.
Installation media is created on a USB drive with liveusb-creator. To boot the installer, insert the liveusb drive and hit esc after powering on. This brings up the boot menu where the usb drive can be selected. The install should be mostly defaults, except that disk encryption and a grub password should be enabled.

After an install and reboot, plug into a wired ethernet connection, install the RPM Fusion software repository, do the first big system software update, then install and remove some packages. Note that kmod-wl from RPM Fusion provides the driver for the Broadcom wireless card after a reboot.

# yum -y install yum-fastestmirror
# yum -y update
# yum localinstall --nogpgcheck \
http://download1.rpmfusion.org/free/fedora/rpmfusion-free-release-stable.noarch.rpm \
http://download1.rpmfusion.org/nonfree/fedora/rpmfusion-nonfree-release-stable.noarch.rpm
# yum -y install kmod-wl akmod-wl krb5-workstation tkinter \
python-setuptools ipython mysql-workbench gconf-editor git \
subversion ethtool liveusb-creator telnet freeglut-devel \
gcc gcc-c++ SDL-devel rdesktop thunderbird thunderbird-enigmail \
vim fortune-mod lsb redhat-lsb-graphics qt-x11 powertop dosbox \
tigervnc tigervnc-server audacity-freeworld rdiff-backup \
xscreensaver-extras-gss tempest-gnome-screensaver \
samba-client rtorrent unrar \
openoffice.org-writer openoffice.org-calc openoffice.org-impress \
screen nmap net-snmp-utils wireshark-gnome iptraf strace sysstat \
rfkill mono gimp rpmdevtools yum-utils ncurses-devel \
libpng-devel mednafen jwhois java-1.6.0-openjdk-plugin wget \
ddrescue wodim vlc mplayer mencoder grip lame easytag id3v2 \
lm_sensors xterm xorg-x11-xinit-session xorg-x11-apps
# yum remove alsa-plugins-pulseaudio

The biggest tip with a 10" netbook is to consolidate and auto-hide menus/toolbars for maximum screen real estate. And use a full-sized LCD monitor when sitting at a desk.

Now, one could stop here and be finished, but I'm interested in tweaking for power saving features, like the Super Hybrid Engine (SuperHE) features of this model. Currently, the acpi_osi and acpi_backlight kernel options must be added to get the /sys/devices/platform/eeepc/cpufv device functional. Add the kernel options in /boot/grub/grub.conf:

kernel /vmlinuz-2.6.35.11-83.fc14.i686 ro root=/dev/mapper/vg_foo-lv_root rd_LUKS_UUID=luks-xxx-xxx-xxx rd_LVM_LV=vg_foo/lv_root rd_LVM_LV=vg_foo/lv_swap rd_NO_MD rd_NO_DM LANG=en_US.UTF-8 SYSFONT=latarcyrheb-sun16 KEYBOARDTYPE=pc KEYTABLE=us rhgb quiet elevator=noop acpi_osi=Linux acpi_backlight=vendor

Enable noatime on ext4 file systems, and tmpfs for temp directories in /etc/fstab:

/dev/mapper/vg_foo-lv_root / ext4 defaults,noatime 1 1
tmpfs /tmp tmpfs defaults,size=256M 0 0
tmpfs /var/tmp tmpfs defaults,size=64M 0 0
tmpfs /var/log tmpfs defaults,size=32M 0 0

Add some power tweak commands, mostly from powertop, to /etc/rc.d/rc.local

### some power and ssd saving tweaks
echo noop > /sys/block/sda/queue/scheduler 
echo 0 > /proc/sys/vm/swappiness
echo 50 > /proc/sys/vm/vfs_cache_pressure
echo 1500 > /proc/sys/vm/dirty_writeback_centisecs
echo 20 > /proc/sys/vm/dirty_ratio
echo 10 > /proc/sys/vm/dirty_background_ratio
echo 1 > /sys/devices/system/cpu/sched_smt_power_savings
echo 10 > /sys/module/snd_hda_intel/parameters/power_save
echo 5 > /proc/sys/vm/laptop_mode
echo ondemand > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
echo ondemand > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
echo ondemand > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor
echo ondemand > /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor
cat /sys/devices/system/cpu/cpu0/cpufreq/ondemand/sampling_rate_max > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/sampling_rate
echo min_power > /sys/class/scsi_host/host0/link_power_management_policy
ethtool -s eth0 wol d
for I in `find /sys -name autosuspend -exec echo {} \;` ; do echo "0" > "$I" ; done
/sbin/rfkill block bluetooth
echo 0 > /proc/sys/kernel/nmi_watchdog

Install the Jupiter package with Eee PC support to automatically tweak many of the above settings, including SuperHE.

Disable some unused services:

# for s in abrtd atd auditd avahi-daemon ip6tables iscsi iscsid mdmonitor portreserve livesys livesys-late; do echo "chkconfig $s off"; chkconfig $s off; done

If restoring a home directory from backup, with SELinux enabled, make sure to restore the security context of the home directories:

# restorecon -vr /home

Google Chrome Beta is an excellent browser for a netbook. It is a very efficient in terms of performance speed and screen usage. To enable the built-in flash plugin, go to the URL about:plugins, expand details, and enable Flash.

Skype is a must for video chat and international voice calling. To get automatic updates, create a repo file for yum in /etc/yum.repos.d/skype.repo:

[skype]
name=Skype Repository
baseurl=http://download.skype.com/linux/repos/fedora/updates/i586/
gpgkey=file:///etc/pki/rpm-gpg/RPM-GPG-KEY-skype

And to get get the latest Adobe Reader, install the adobe repo rpm:

# yum localinstall --nogpgcheck http://linuxdownload.adobe.com/adobe-release/adobe-release-i386-1.0-1.noarch.rpm
# yum -y install AdobeReader_enu

Someone on a mailing list pointed out that the "Windows" key can be made more useful by mapping it as a main menu key:

# xmodmap -e 'remove mod4 = Super_L' 
# gconftool-2 --set /apps/metacity/global_keybindings/panel_main_menu --type=string 'Super_L'

The trackpad is nice, supporting one, two and three-finger tapping. Enable this in the Gnome mouse preferences, for easy copy-and-paste the way it was meant to be. On Linux, mouse button one is highlight/select, copy being automatic, and button three is paste. With a trackpad this is double-tap and drag one finger to select text, then a three-finger single-tap to paste.

To update the bios, here are my notes:

update bios 1015PE.ROM on fat16 drive
http://support.asus.com/download/
unzip 1015PE-ASUS-0904.zip
insert a free usb memory stick, assuming it is sdb...
umount /dev/sdb1
parted -s -- /dev/sdb mklabel msdos
parted -s -- /dev/sdb mkpartfs primary fat16 0 32
mount -t vfat -o fat=16 /dev/sdb1 /mnt
cp 1015PE-ASUS-0904.ROM /mnt/1015PE.ROM
umount /dev/sdb1
reboot with the usb drive & power plugged in
hit alt+f2 first thing at post
if all is well, it will get past reading 1015PE.ROM
allow the update to complete! ~10 minutes

You can change video modes from the command line. Stick these into ~/.bashrc:

alias vga='xrandr --fb 1280x1624 --output LVDS1 --primary --mode 1024x600 --rate 60 --pos 0x0 --panning 0x0 --output VGA1 --mode 1280x1024 --rate 60 --pos 0x1024 --above LVDS1'
alias mir='xrandr --fb 1280x1024 --output LVDS1 --primary --mode 1024x600 --rate 60 --pos 0x0 --panning 1280x1024 --output VGA1 --mode 1280x1024 --rate 60 --pos 0x0 --same-as LVDS1'
alias lcd='xrandr --fb 1024x600  --output LVDS1 --primary --mode 1024x600 --rate 60 --output VGA1 --off'

Other things done:

Recover LDAP Auth on Mac OS X

2011-02-21T17:01:00.002-05:00

I have an entire rack of Apple Xserve nodes, running Mac OS X 10.5. The Xserve is fond of refusing to reboot, crashing, and corrupting the Open Directory LDAP database when under high load. Here is a quick fix:

$ su -
# tar cvzf ldap-YYYYMMDD.tar.gz /var/db/openldap
# rm /var/db/openldap/openldap-data/log.0*
# launchctl unload /System/Library/LaunchDaemons/org.openldap.slapd.plist
# /usr/bin/db_recover -h /var/db/openldap/openldap-data
# launchctl load /System/Library/LaunchDaemons/org.openldap.slapd.plist
# slapcat | less

ATLAS NumPy SciPy Build on RHEL 6

2011-02-21T14:39:00.002-05:00

This is a build log for the latest ATLAS (ATLAS/BLAS/LAPACK), NumPy, and SciPy on Red Hat Enterprise Linux 6. Please note that this was attempted many times, following the documentation, but there were still errors when importing certain functions. The big tip is to edit site.cfg before building and installing numpy.
Here is an example of some of the errors we would see with a mis-configured build:

from scipy.optimize import fsolve
import scipy.integrate as integrate
from scipy.integrate import odeint
ImportError: /usr/local/scipy/lib64/python2.6/site-packages/scipy/optimize/_lbfgsb.so: undefined symbol: ATL_dcopy
ImportError: /usr/local/scipy/lib64/python2.6/site-packages/scipy/integrate/_odepack.so: undefined symbol: ATL_idamax
ImportError: /usr/local/scipy/lib64/python2.6/site-packages/scipy/integrate/_odepack.so: undefined symbol: daxpy_

Here is a log of commands for a successful build. We install each package into its own version-numbered directory, soft-linking to the current version and then setting the user environment variables.

# cd /usr/local/src
# wget http://www.netlib.org/lapack/lapack-3.3.0.tgz
# tar xzvf lapack-3.3.0.tgz
# cd lapack-3.3.0
# cp INSTALL/make.inc.gfortran make.inc
# vim make.inc
add -fPIC to OPTS and NOOPT
# make 2>&1 | tee log.make
# make blaslib lapacklib tmglib lapack_testing 2>&1 | tee log.makeall

# cd /usr/local/src
# wget http://downloads.sourceforge.net/project/math-atlas/Developer%20%28unstable%29/3.9.35/atlas3.9.35.tar.bz2
# tar xjvf atlas3.9.35.tar.bz2
# mv ATLAS atlas-3.9.35
# cd atlas-3.9.35
# mkdir ATLAS_LINUX ; cd ATLAS_LINUX
# cpufreq-selector -g performance
# ../configure -Fa alg -fPIC -Si cputhrchk 0 --prefix=/usr/local/atlas-3.9.35 --with-netlib-lapack-tarfile=/usr/local/src/lapack-3.3.0.tgz 2>&1 | tee log.config
# make 2>&1 | tee log.make
THIS WILL TAKE HOURS TO MAKE!
# make install 2>&1 | tee log.install
# cd /usr/local ; ln -s atlas-3.9.35 atlas
# export LD_LIBRARY_PATH=/usr/local/atlas/lib:$LD_LIBRARY_PATH

# cd /usr/local/src
# wget http://cdnetworks-us-2.dl.sourceforge.net/project/numpy/NumPy/1.5.1/numpy-1.5.1.tar.gz
# tar xzvf numpy-1.5.1.tar.gz
# cd numpy-1.5.1
# cp site.cfg.example site.cfg
# vim site.cfg
[DEFAULT]
library_dirs = /usr/local/atlas/lib
include_dirs = /usr/local/atlas/include
[atlas]
atlas_libs = lapack, f77blas, cblas, atlas
# python setup.py build 2>&1 | tee log.build
# python setup.py install --prefix=/usr/local/numpy-1.5.1 2>&1 | tee log.install
# cd /usr/local ; ln -s numpy-1.5.1 numpy
# export PYTHONPATH=$PYTHONPATH:/usr/local/numpy/lib64/python2.6/site-packages
# export PATH=/usr/local/numpy/bin:$PATH
# python /usr/local/numpy/lib64/python2.6/site-packages/numpy/distutils/system_info.py

# cd /usr/local/src
# wget http://cdnetworks-us-2.dl.sourceforge.net/project/scipy/scipy/0.9.0rc3/scipy-0.9.0rc3.tar.gz
# tar xzvf scipy-0.9.0rc3.tar.gz
# cd scipy-0.9.0rc3
# python setup.py build 2>&1 | tee log.build
# python setup.py install --prefix=/usr/local/scipy-0.9.0rc3 2>&1 | tee log.install
# cd /usr/local ; ln -s scipy-0.9.0rc3 scipy

Some references: