Build CouchDB is a wrapper or master project which pulls in, from official sources, CouchDB plus all of its dependencies. It is the most straightforward and reliable procedure to build official CouchDB releases from source.
Build CouchDB builds an isolated, independent server. You do not need administrator access to run it. You can run several couches (for example, 0.10, 0.11, 1.0 releases) side-by-side.
Build CouchDB is developed and tested on the following operating systems:
- Red Hat Enterprise Linux Server release 5.5 (Tikanga)
- CentOS 5.5
- Debian GNU/Linux 5.0 (Lenny)
- 9.10 (Karmic Koala)
- 10.04 LTS (Lucid Lynx)
- 10.10 (Maverik Meerkat)
- 11.04 (Natty Narwhal)
- 11.10 (Oneiric Ocelot)
- Fedora 13
- Mac OS X
- OpenSUSE 11.3
- Scientific Linux 5.3
- Solaris 10, OpenSolaris
The following systems are planned for support in the near future:
- MS Windows Vista, Windows 7
You need only a few packages provided by the operating system. Copy and paste the commands below.
sudo yum install gcc gcc-c++ libtool \ zlib-devel openssl-devel rubygem-rake
On Red Hat Enterprise Linux:
The procedure is the same as Fedora, but also install the
On Debian, first install
sudo and add yourself to
su - apt-get install sudo visudo
On Ubuntu and Debian:
sudo apt-get install make gcc zlib1g-dev libssl-dev rake
sudo zypper install flex lksctp-tools-devel zip \ rubygem-rake gcc-c++ make m4 zlib-devel \ libopenssl-devel libtool automake
On Scientific Linux
sudo yum install --enablerepo=dag gcc gcc-c++ libtool zlib-devel openssl-devel \ autoconf213
This tool only supports the OpenCSW toolchain. If you have success with other tools, let me know!
The Solaris Studio (formerly SunStudio) tools are required. On Solaris 10, it's easy.
sudo pkg install ss-dev
On Solaris 11, go to the Solaris Studio download page and get the Package Installer file you need. Copy it to your Solaris machine and follow the Solaris Studio installation instructions, something like this:
bzcat SolarisStudio12.2-*.tar.bz2 | /bin/tar -xf - cd SolarisStudio-12.2-*-pkg-ML
Also, OpenCSW packages are needed.
pkgadd -d http://mirror.opencsw.org/opencsw/pkg_get.pkg # Answer all questions affirmatively
Add CSW to your path. This must always be in the PATH. Every time you log
in, you must set the correct
$PATH (or make it automatic in
Change the package archive (ibiblio URL is down) by running
vi /opt/csw/etc/pkg-get.conf and setting
url=ftp://ftp.ibiblio.org/pub/mirrors/opencsw/current. Save and exit, then
Finally, install Rake from OpenCSW:
sudo pkg-get install ruby rake # Also perhaps "git"
On Mac OS X, install Xcode.
Getting the Code
You will need the Git tool. Check out the code and pull in the third-party submodules.
git clone git://github.com/iriscouch/build-couchdb cd build-couchdb git submodule init git submodule update
How to Build CouchDB
Just run Rake.
CouchDB and all its dependencies will install in the
build/. To uninstall,
simply delete that directory.
It's CouchDB! Just type
couchdb. (But remember the path)
$ build/bin/couchdb Apache CouchDB 0.12.0aa63efb6-git (LogLevel=info) is starting. Apache CouchDB has started. Time to relax. [info] [<0.33.0>] Apache CouchDB has started on http://127.0.0.1:5984/
You can of course call it by absolute path. If your username is
amit and you
checked out the code in your home directory, you would run:
Conveniently Add CouchDB to the PATH
The build process creates a small shell script,
build/env.sh. The script
will add the buid to your shell's
$PATH. This will only affect that shell
session, other terminals or shell sessions will not change. (This is on
purpose, to isolate CouchDB, so that it is easy to remove, or so multiple
versions can be installed side-by-side.)
Simply source the script when you want to use CouchDB.
Your working directory needn't be anywhere special when sourcing the file.
It can be processed from anywhere. The idea is, when you are working, you
realize you need couchdb, just type
. ~/my/stuff/code/build-couchdb/build/env.sh or whatever and it will work.
You can source the file as often as you like. Subsequent exection will not do anything.
. build/env.sh . build/env.sh # Sourcing with wild abandon!
If the file is read from a script or in a pipeline, it will execute silently (by detecting whether it is connected to a TTY terminal).
Build CouchDB supports some fancy tricks by entering cheat codes to the Rake command.
Build any Git branch or tag of Erlang/OTP
erl_checkout parameter with the commit id, branch name, or tag name.
For example, to build with Erlang R13B04 release:
Build any Git branch or tag of CouchDB
git parameter with the repository URL, then a space, then the branch,
tag, or commit hash. (Remember to quote all of those so Rake sees the space.)
Want to build GeoCouch? No problem.
rake git="git://github.com/vmx/couchdb geocouch"
Any CouchDB plugin can be loaded remotely from Git, built, and installed into the final CouchDB system.
rake plugin="git://github.com/vmx/couchdb origin/gc-separate"
Multiple plugins can be processed together:
rake plugins="git://github.com/vmx/couchdb origin/gc-separate,git://github.com/somebody/whatever some_tag"
plugins supports comma-separated lists; use whichever
you remember better.)
Install CouchDB somewhere besides
install parameter to place the final couchdb binaries anywhere.
Build CouchDB makes it simple to install several couchdb versions side-by-side.
rake git="git://github.com/vmx/couchdb geocouch" install=geocouch rake git="git://git.apache.org/couchdb.git trunk" install=trunk for tag in 1.0.1 11.0 11.1; do rake git="git://git.apache.org/couchdb.git tags/$tag" install=$tag done
install needs to be an absolute path. Use:
e.g. creates the directory
geocouch in your current directory.
For side-by-side installs there is a small shortcut to avoid rebuilding Erlang:
couchdb_build variable instead, which will install CouchDB separately
from its dependencies. Just remember never to move or delete the dependencies!
rake install=/dependencies/go/here couchdb_build=/but/couch/goes/here
Get a manifest of all the components
To get a better idea of exactly what is going on, add a
That will produce additional files in
build/manifest which indicate which
package (icu, erlang, spidermonkey, etc) owns which files within
trick I do a lot is
cat build/manifest/couchdb | xargs rm to "uninstall" only
couchdb so I can try a rebuild.
I have no idea how
manifest interacts with
install as I have never
used them together.
Do not strip down Erlang/OTP
Build CouchDB strips many modules out of the Erlang platform to reduce disk
usage. (You can see which ones at the top of
tasks/erlang.rake.) To indicate
that a package should be kept, set the
otp_keep variable to space-separated
rake otp_keep="compiler eunit"
Or, you can keep them all this way:
How to build only Erlang, couchjs, and OTP so you can build your own CouchDB elsewhere
There is a special shortcut task to build everything CouchDB needs (i.e. its dependencies).
rake couchdb:deps otp_keep="*"
Be careful not to build the
couchdb target because after it completes, it will delete Erlang components needed for building (but not running).
Next, there is a simple task which outputs a
sh script used to configure any CouchDB checkout.
rake --silent environment:configure
The output will look similar to this:
export PATH="/Users/jhs/src/build-couchdb/build/bin:$PATH" LDFLAGS='-R/Users/jhs/src/build-couchdb/build/lib -L/Users/jhs/src/build-couchdb/build/lib' CFLAGS='-I/Users/jhs/src/build-couchdb/build/include/js -I/Users/jhs/src/build-couchdb/build/lib/erlang/usr/include' ./configure
In the CouchDB source, paste the above code after running
./bootstrap. Next, you can run
make dev, or anything.