tahoe-lafs/README
Zooko O'Whielacronx bc60e57cce more README updates in response to user feedback
The user in this case is my brother Josh.
2007-04-27 16:03:51 -07:00

186 lines
7.8 KiB
Plaintext

Welcome to the AllMyData "tahoe" project. This project implements a
secure, distributed, fault-tolerant storage mesh.
The basic idea is that the data in this storage mesh is spread over all
participating nodes, using an algorithm that can recover the data even if a
majority of the nodes are no longer available.
The main application code is in the 'allmydata' package, under src/allmydata/
. There is also a patched version of PyCrypto (adding a faster CTR-mode) in
src/Crypto/ which gets installed to the 'allmydata.Crypto' package (since the
API is different than the normal Crypto package). It also includes Zooko's
PyFEC library, a fast python wrapper around the Rizzo 'fec' C library,
installed to the 'zfec' package and located in src/zfec/ .
LICENCE:
Tahoe is offered under the GNU General Public License (v2 or later), with
the added permission that, if you become obligated to release a derived work
under this licence (as per section 2.b), you may delay the fulfillment of
this obligation for up to 12 months. See the COPYING file for details.
DEPENDENCIES:
Note: Except for Foolscap, all of the following dependencies can
probably be installed through your standard package management tool
if you are running on a modern Unix operating system. If you are
running any modern Linux or *BSD distribution then you can almost
certainly get them through your standard package manager. If you are
running Mac OS X then the "fink" package management tool does not
have most of these packages, but the "darwinports" package management
tool appears to have them. If you are running on Windows then I'm
afraid you'll have to install them by hand (although the "cygwin"
package management tool does have some of them). If you are running
on Solaris, I would like to hear from you -- I have no idea how it is
done on Solaris nowadays.
* a C compiler (language)
* GNU make (build tool)
* Python 2.4 or newer (tested against 2.4, 2.5, and 2.5.1), but v2.5 or
higher is required on Windows-native) (language)
http://python.org/
* Python Twisted (tested against both 2.4 and 2.5) (network and operating
system integration library)
http://twistedmatrix.com/
You need the following subpackages (which are included in the
default Twisted distribution):
* core (the standard Twisted package)
* web, trial, conch
Twisted requires zope.interface, a copy of which is included in the Twisted
distribution.
* Python Nevow (probably 0.9.0 or later) (web presentation language)
http://divmod.org/trac/wiki/DivmodNevow
* Python setuptools (build and distribution tool)
http://peak.telecommunity.com/DevCenter/EasyInstall#installation-instructions
* Python PyOpenSSL (0.6 or later) (secure transport layer)
http://pyopenssl.sourceforge.net
To install PyOpenSSL on Windows-native, download this:
http://allmydata.org/source/pyOpenSSL-0.6.win32-py2.5.exe
* to build the debian packages you will need all the usual debian-packaging
tools, which means the 'build-essential' metapackage and all of the
packages listed as "Build-Depends" in DIST/debian/control for your
distribution. You will also want the 'fakeroot' package to allow the
top-level 'make deb-DIST' targets work.
* Foolscap (0.1.2 or newer) (remote object library)
http://twistedmatrix.com/trac/wiki/FoolsCap
Note: since the Foolscap wire protocol is not yet compatible from one
release to the next, make sure all of your nodes are using the same version
of Foolscap
* on Windows, the pywin32 package
http://sourceforge.net/projects/pywin32/
BUILDING:
Just type 'make'. This works on Windows too, provided that you have the
dependencies mentioned above (either a normal cygwin build or a mingw-style
native build is supported by the makefile -- the cygwin build is the
default).
If the desired version of 'python' is not already on your PATH, then type
'make PYTHON=/path/to/your/preferred/python'.
'make test' runs the unit test suite.
INSTALLING:
If you're running on a debian system, use 'make deb-dapper' or 'make
deb-sid' to construct a debian package named 'allmydata-tahoe', which you
can then install.
If not, you'll need to run three separate install steps, one for each of the
three subpackages (allmydata, allmydata.Crypto, and zfec). You may wish to
use a different version of 'python' for these steps, or provide a --prefix
or --root argument for the install.
cd src/zfec && python setup.py install && cd ../..
cd src/Crypto && python setup.py install && cd ../..
# the allmydata subpackage's setup.py script is in the root directory
python setup.py install
To test that all the modules got installed properly, start a python
interpreter and import modules as follows:
% python
Python 2.4.4 (#2, Jan 13 2007, 17:50:26)
[GCC 4.1.2 20061115 (prerelease) (Debian 4.1.1-21)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import allmydata.Crypto
>>> import allmydata.interfaces
>>> import fec
>>>
To run from a source tree (without installing first), type 'make', which
will put all the necessary libraries into a local directory named
instdir/lib/pythonN.N/site-packages/ , which you can then add to your
PYTHONPATH .
RUNNING:
If you installed one of the debian packages constructed by "make deb-*" then
it creates an 'allmydata-tahoe' executable, usually in /usr/bin . If you
didn't install a package you can find allmydata-tahoe in bin/ . This tool is
used to create, start, and stop nodes. Each node lives in a separate base
directory, inside of which you can add files to configure and control the
node. Nodes also read and write files within that directory.
A mesh consists of a single central 'introducer' node and a large number of
'client' nodes. If you are joining an existing mesh, the introducer node
will already be running, and you'll just need to create a client node. If
you're creating a brand new mesh, you'll need to create both an introducer
and a client (and then invite other people to create their own client nodes
and join your mesh).
The introducer node is constructed by running 'allmydata-tahoe
create-introducer --basedir $HERE'. Once constructed, you can start the
introducer by running 'allmydata-tahoe start --basedir $HERE' (or, if you
are already in the introducer's base directory, just type 'allmydata-tahoe
start'). Inside that base directory, there will be a pair of files
'introducer.furl' and 'vdrive.furl'. Make a copy of these, as they'll be
needed on the client nodes.
To construct a client node, pick a new working directory for it, then run
'allmydata-tahoe create-client --basedir $HERE'. Copy the two .furl files
from the introducer into this new directory, then run 'allmydata-tahoe start
--basedir $HERE'. After that, the client node should be off and running. The
first thing it will do is connect to the introducer and introduce itself to
all other nodes on the mesh. You can follow its progress by looking at the
$HERE/twistd.log file.
To actually use the client, enable the web interface by writing a port
number (like "8080") into a file named $HERE/webport and then restarting the
node with 'allmydata-tahoe restart --basedir $HERE'. This will prompt the
client node to run a webserver on the desired port, through which you can
view, upload, download, and delete files.
A client node directory can also be created without installing the code
first. Just use 'make create-client', and a new directory named 'CLIENTDIR'
will be created inside the top of the source tree. Copy the relevant .furl
files in, set the webport, then start the node by using 'make start-client'.
To stop it again, use 'make stop-client'. Similar makefile targets exist for
making and running an introducer node.
There is a public mesh available for testing. Look at the wiki page
(http://allmydata.org) for the necessary .furl data.