#/bin/sh 

VERSIONS=(5 5.1)
REPOS=(os updates centosplus extras)
RSYNC_LOCATION="rsync://rsync.mirrorservice.org/sites/mirror.centos.org"
DESTINATION="/repository/distributions/CentOS" 

for version in ${VERSIONS[@]}
do
        for repo in ${REPOS[@]}
        do
                echo "*******************************************************"
                echo "      rsyncing CentOS ${version} ${repo} repository  "
                echo "*******************************************************" 

                echo "Source: ${RSYNC_LOCATION}/${version}/${repo}/i386/"
                echo "Destination: ${DESTINATION}/${version}/${repo}/i386/" 

                /usr/bin/rsync -q -Pvptrl --delete ${RSYNC_LOCATION}/${version}/${repo}/i386/ ${DESTINATION}/${version}/${repo}/i386/
        done
done

You should be able to run this from cron quite easily, don’t forget to point to a mirror near you!

Firstly make sure you have all the Python Eggs -

easy_install Pylons==0.6.1
easy_install SQLAlchemy==0.4.1
easy_install Elixir==0.4

Configuration

The path to your database needs specifying in

# development.ini
sqlalchemy.default.url = sqlite:///%(here)s/databases/mydb.db

Application Environment Configuration

This reads in the configuration specified above into the global configuration space of the application and binds the session (that are used to perform sql operations with) and the metadata (that maps your object to table entity relations) to the actual physical database. The session and metadata are defined in the model below.

# project/config/environment.py
# at the end of the file add
from sqlalchemy import engine_from_config
config['pylons.g'].default_engine = engine_from_config(config, 'sqlalchemy.default.')
from project import model
model.session.bind = config['pylons.g'].default_engine
model.metadata.bind = config['pylons.g'].default_engine

Model Setup

Assuming you define your model in project/model/mymodel.py as follows -

# project/model/mymodel.py
from sqlalchemy.orm import scoped_session, sessionmaker
from sqlalchemy import create_engine
from elixir import *

class A(Entity):
    ...

Then your project/model/__init__.py file should contain -

# project/model/__init__.py
from mymodel import *
setup_all()

The setup_all() call tells elixir to build the mappings between the defined objects (that inherit from Entity) and the table definitions. The bind call in environment.py makes these table definitions meaningful!

Application Setup

If you want to run “paster setup-app development.ini” to create the database initially edit project/websetup.py. The global environment is not imported into websetup because it’s a one time only operation, so it needs to be specified again.

# project/websetup.py
# at the end of the file
from provisionz import model
print "Creating tables"
config['pylons.g'].default_engine.echo = True
model.metadata.bind = config['pylons.g'].default_engine
model.drop_all()
model.create_all()

If you plan on adding some initial data to the database, then a few more things need to be added to the above

model.session.bind = config['pylons.g'].default_engine
mytable = model.MyTable(column="my data")
model.session.flush()

Hopefully this is self explanitory, the session is used only to query the database. The metadata is used to create it but it must exist (and be bound) for the session to work.

Application Globals and Cleanup

The file project/lib/base.py contains the BaseController method which is inherited by all controllers you will define, it’s also a good place to add your global model import, along with any other imports -

from project import model

And at the end of the BaseController __call__ method make sure it finishes by cleaning up the session connection -

try:
    return WSGIController.__call__(self, environ, start_response)
finally:
    model.session.remove()

Now in the controllers you have defined you can use and abuse your model -

row = model.MyTable(...)
model.session.flush()

It’s very straight forward on Windows or Linux, the one thing you do need though is some other computer running an SSH server on the internet to connect to. This could be a home based PC or just your OpenWRT ADSL router, or as fancy as dedicated server or ssh service, it’s up to you.

If your running Windows, download PuTTY and configure it to tunnel a local port, say 5000, using a Dynamic connection to your remote SSH server.

On Linux with OpenSSH the same command to do this is -

ssh -f -N -D 5000 remoteuser@remote.host.com

The -f and -N make backgrounding the process possible in a terminal, it’s the -D that is the import part.

Then just configure your browser, in Firefox it’s under Edit -> Preference and Windows it’s Tools -> Options; click on the Advanced tab, then Network sub tab and finally select the Settings button. At the bottom of the list is the SOCKS 5 option. In SOCKS host put “localhost” and under the port “5000″.

Thats it, browse away. If you do this often, make sure you try out one of the Proxy Selection Firefox add-ons.

Disclaimer: Your DNS traffic will not tunneled or encrypted but all your web traffic is using this method.

ssh -f -N -L 110:localhost:110 ruser@remote.system.com

Then you would configure your POP mail client to connect to localhost:110 instead of the remote system.

Other times you may want to make a connection from a remote machine to your local machine (a reverse tunnel). A typical example of this is when connecting from a home pc to a work system that is behind a firewall. To setup this sort of tunnel the tunnel would be initiated on the work system as such:

ssh -fnNT -R 5678:localhost:22 home.system.com

Then it will be possible to login to your work pc by running ssh localhost -p 5678 on your home pc.

However, what if you need to tie these two things together by having an intemediary server? If we take the example this time of running vnc over ssh so we want something like this:

VNC-server (5901) <-- ssh --> (5901) In-between-system (5901) <-- ssh --> (5901) VNC-client

How do we do it? Firstly, startup the vnc server on the VNC-server. Then startup the reverse tunnel from the VNC-server to the In-between-system like this:

ssh -ngfNT -R 5901:localhost:5901 ruser@in.between.system

Note: For the above process to background itself correctly as the “-f” command implies you would need your keys setup between these systems (i.e. the public key of the VNC-server copied to the authorized_keys file of “ruser” on the In-between-system).

Then on your VNC client start a local tunnel to the In-between-system as follows:

ssh -f -N -L 5901:localhost:5901 ruser@in.between.system

Then to view your remote desktop run vncviewer localhost:1 (:1 representing port 5901) which will route your data via the In-between-system and tunnel it over ssh.

Then the minimum that’s required is -

cd rpm/SPECS
rpmbuild -ba --target=i686 kernel-2.6.spec

If you need to patch the kernel with your changes, place the patch in the rpm/SOURCES directory and update the spec file with the appropriate lines. For example, i needed to add a patch for a Brain Boxes serial card.

I added a line as follows in the header section:

Patch1024: linux-2.6-brainboxes.patch

and then further down in the %prep where the patching actually takes place, i added:

%patch1024 -p1

As i had other third party modules built against this kernel i maintained the existing version and release number, however it would be a good idea to change this if you are releasing it into the wild!

The kernel kbuild documenation does not have a complete description of this, so i had to piece it together from mailing lists. The reason for this is probably simply that doing this is discouraged because drivers should have their code released and if possible integrated into the kernel.

So, you’ve got some object files that make up a module but need to link (ld) them against the current kernel. It is NOT as simple as doing ld -o mymod.ko *.o -lc -L/usr/src/linux.

So i have these modules:

obj1.o
obj2.o
obj3.o

And i want them to turn end up as mymod.ko. The first trick is to rename them to name.o_shipped. I say trick because the kbuild system does a lot of stuff that is not shown in the makefile. It is invoked by the obj-m definition in the makefile below (see comments).

for i in *.o; do
    mv $i $i_shipped
done

ls
obj1.o_shipped obj2.o_shipped obj3.o_shipped

I now need to create a Makefile that works correctly with the kbuild system, and uses these object files:

MODULE_NAME := mymod
MODULE_OBJECT := $(MODULE_NAME).ko

# these binary objects must be renamed to 

The important bit here is the line $(MODULE_NAME)-y := $(CORE_OBJS) which tells the kbuild system to look for files ending with “_shipped”.

I hope this helps somebody.

Discaimer: this may not be 100% correct as i’m no kernel expert but it works for me!

Additionally the create_devices.sh script mentioned in the documentation didn’t work for me (when using version 1.5.24-2).

All i wanted to do was dd a disk image file (specifically a Linux boot disk image) to my usb flash drive so i could use it for kicking off network installations. Simple you might think but of no (unless you read this post first that is!).

First of all i though, ah i’ll have a look at my device list -

ls -lht /dev

nothing, no devices in fact /dev doesn’t even exist - or so i first thought. It’s simulated internally, so as the documentation says /dev/tty does actually exist and you can run ls against it and it’ll work. A little bit like magic always gets people.

Then i thought, ok try something simple. My device is mounted as o: under windows so i thought maybe /cygdrive/o will be a raw device.

dd if=/path/to/myfile of=/cygdrive/o

error, nope it’s not.

Thats when i started googling and came across a multitude of incorrect methods (this post for one) for mounting and accessing raw devices.

dd if=/path/to/myfile of=//./O:
mount -s -b //./o: /dev/fd0
mount -f -b //./physicaldrive5 /dev/fd0

the last one started me off looking at device numbers which was helpful. Check the Windows XP Computer Manager -> Disk Manager tool to find out what the device number actually is.

One thing i was fairly sure about is that you definately need to mount it within Cygwin before you can actually use it as a raw device. Eventually, i read the documentation thourouly and discovered a post saying how all the old methods are now obsolete.

What you need to do is find out the device number from Disk Manager tool (is there a simpler command line way i wonder?) and then look for the corresponding device in /dev. For example, my device was mounted as O: in Windows but this has absolutely no bearing on the device number so i had to check to find out that it was device 5 (counting from 0) so it was actually the sixth device.

Thus in /dev, /dev/sda equals device 0 which is my C: drive so /dev/sdf is my flash device!

I’ve just found it and it works like a dream, just download the zip file and put the files somewhere in your path. I put them in H:\Documents and Settings\Nick\My Documents\My Programs\puttycyg and ensure that is in my path.

Fire it up (putty.exe that is) and select ‘Cygterm’ and ‘-’ for the command line and it just works!

I’m learning Pylons which is an MVC framework at the moment and running through the blog tutorial which is a great starting point. Im developing using an Sqlite database (and probably deploying on a Mysql database), using the SQLObject [model] component of Pylons.

I came a bit of a cropper and wasted a few hours trying to discover the correct URI syntax for SQLObject with sqlite on Windows. On linux typically it’s easy but i couldn’t let it rest and had to figure it out how do it on Windows so i could develop project easily and quickly on whichever platform i want to develop on. I’m adding this entry here so i never have to figure it out again!

I created a database with SQLiteSpy (great SQLite GUI tool) and copied the .db3 file to a directory called H:\dbs.

The SQLObject scheme syntax for this is as follows, which i entered into the development.ini Pylons file:

sqlobject.dburi = sqlite:///H|/dbs/blogtutorial.db3

To get around this is i just moved the device file in the sysimage location for the duration of the yum install process. It’s a hack but it works and thats the main thing.

mv /dev/log /dev/log.tmp
yum install <package>
mv /dev/log.tmp /dev/log