It is recommended that you create one or more dedicated virtual machines to try out and develop Cockpit.
While playing with Cockpit you will very likely want to run experimental code that does dangerous things, such as formatting block devices. And even if the code has no bugs, you might want to intentionally do destructive things that you would never want to do to a production system. And last but not least, it is easier to add more virtual hardware to a virtual machine for testing, such as more hard disks or network adapters.
Currently the x86_64 architectures of Fedora 20 and Rawhide are most often used for development.
Check tools/cockpit.spec for the concrete Fedora build dependencies.
The following should work in a fresh Git clone:
$ srpm=$(tools/make-srpm)
$ sudo yum-builddep $srpm
$ sudo yum install nodejs npm
In addition for testing the following dependencies are required:
$ sudo yum install trickle nbd-server python-libguestfs qemu \
mock qemu-kvm python curl libvirt-client libvirtd qemu-nbd \
krb5-workstation krb5-server
$ npm install phantomjs
Cockpit uses the autotools and thus there are the familiar ./configure
script and the familar Makefile targets.
But after a fresh clone of the Cockpit sources, you need to prepare them by running autogen.sh. Maybe like so:
$ mkdir build
$ cd build
$ ../autogen.sh --prefix=/usr --enable-maintainer-mode --enable-debug
As shown, autogen.sh also runs 'configure' with the given options, and it also prepares the build tree by downloading various nodejs dependencies.
When working with a Git clone, it is therefore best to simply always
run ../autogen.sh instead of ../configure.
Creating a build directory puts the output of the build in a separate directory, rather than mixing it in with the sources, which is confusing.
Then you can build the sources and install them, as usual:
$ make
$ sudo make install
$ sudo cp ../src/bridge/cockpit.pam.insecure /etc/pam.d/cockpit
$ sudo sh -c "cat ../src/bridge/sshd-reauthorize.pam >> /etc/pam.d/sshd"
This will install Cockpit and all support files, and will install a simplistic PAM configuration.
Cockpit has a single non-recursive Makefile. You can only run make
from the top-level and it will always rebuild the whole project.
If you prefer to install to a different --prefix and would prefer
that make install not write outside that prefix, then specify the
--enable-prefix-only option to autogen.sh. This will result in an
installation of Cockpit that does not work without further tweaking.
For advanced users only.
You can run unit tests of the current checkout:
$ make check
These should finish very quickly and it is good practice to do it often.
To run the integration tests, see test/README.
Once Cockpit has been installed, the normal way to run it is via systemd:
# systemctl start cockpit.socket
This will cause systemd to listen on port 9090 and start cockpit-ws when someone connects to it. Cockpit-ws will in turn activate cockpitd via D-Bus when someone logs in successfully.
To run Cockpit without systemd, start the cockpit-ws daemon manually:
# /usr/libexec/cockpit-ws
Then you can connect to port 9090 of the virtual machine. You might need to open the firewall for it. On Fedora:
# firewall-cmd --reload
# firewall-cmd --add-service=cockpit
# firewall-cmd --add-service=cockpit --permanent
Point your browser to https://IP-OR-NAME-OF-YOUR-VM:9090
and Cockpit should load after confirming the self-signed certificate. Log in as root with the normal root password (of the virtual machine).
Cockpit consists of the systemd service: cockpit.service . After installing a new version, you should usually restart it:
# systemctl restart cockpit
and then reload the browser.
If you want to run /usr/libexec/cockpit-ws outside of systemd, stop
it first, including the socket:
# systemctl stop cockpit.socket cockpit
Simple version. Edit the appropriate sources and then:
$ make
$ sudo make install
$ sudo systemctl restart cockpit
Then refresh in your browser and your change should be visible. Note that for pure javascript changes you probably don't need to do the last part.
Bigger changes need to be discussed on #cockpit or our mailing list cockpit-devel@lists.fedoraproject.org before you invest too much time and energy.
Cockpit is a designed project. Anything that the user will see should have design done first. This is done on the wiki and mailing list.
You need to be familiar with git to contribute a change. Do your changes on a branch. Your change should be one or more git commits that each contain one single logical simple reviewable change, without modifications that are unrelated to the commit message.
Make a pull request on github.com with your change. All changes get reviewed, tested and iterated on before getting into Cockpit. Don't feel bad if there's multiple steps back and forth asking for changes or tweaks before your change gets in.
The Cockpit UI code is comprised of HTML, javascript and CSS. Almost all of this code is found in the packages in the pkg/ subdirectory of the Cockpit code.
You can setup a system for rapid development of Cockpit UI code. This will allow you to simply refresh your browser and see any changes you've made to code in the pkg/ subdirectory.
Run this command from your top level Cockpit checkout directory, and make sure to run it as the user that you will using to log into Cockpit.
$ mkdir -p ~/.local/share/cockpit
$ ln -s $(pwd)/pkg/* ~/.local/share/cockpit
This will cause cockpit to read UI files directly from the Cockpit code pkg/ directory instead of using the installed Cockpit UI files. But only for the user which ran the above command.
To revert the above change, run:
$ rm ~/.local/share/cockpit
All messages from the various cockpit services go to the journal and can be seen with commands like:
$ sudo journalctl -f
Much of Cockpit has more verbose internal debug logging that can be enabled when trying to track down a problem. To turn it on add a file to your system like this:
$ sudo mkdir -p /etc/systemd/system/cockpit.service.d
$ sudo printf "[Service]\nEnvironment=G_MESSAGES_DEBUG=cockpit-ws,cockpit-daemon,cockpit-bridge\nUser=root\nGroup=\n" > /etc/systemd/system/cockpit.service.d/debug.conf
$ sudo systemctl daemon-reload
$ sudo systemctl restart cockpit
In the above command you'll notice the string "cockpit-ws". This is a log domain. There are various log domains you can enable:
- cockpit-bridge: Cockpit bridge detailed debug messages
- cockpit-daemon: Cockpit DBus daemon detailed debug messages
- cockpit-protocol: Very verbose low level traffic logging
- cockpit-ws: Cockpit Web Service detailed debug messages
- WebSocket: Verbose low level WebSocket logging
To revert the above logging changes:
$ sudo rm /etc/systemd/system/cockpit.service.d/debug.conf
$ sudo systemctl daemon-reload
$ sudo systemctl restart cockpit
Cockpit communicates with the system via a WebSocket. To log all communication to the Web Browser's console, run one of the following commands in the console:
> window.debugging = "channel"
Or in order to log starting at page reload:
> window.sessionStorage["debugging"] = "channel"
You may want to run cockpit-ws under a debugger such as valgrind or gdb. You can run these processes as your own user, although you won't be able to debug all the authentication logic in those cases.
First of all make sure Cockpit is installed correctly. Even though we will be running cockpit-ws from the built sources this still relies on some of the right bits being installed in order for Cockpit to work (ie: PAM stack, UI files, cockpit-bridge, etc.)
This is how you would run cockpit-ws under gdb:
$ export G_DEBUG=fatal-criticals
$ export G_MESSAGES_DEBUG=cockpit-ws,cockpit-daemon,cockpit-bridge
$ gdb --args ./cockpit-ws --port 10000 --no-tls
And you can run cockpit-ws and cockpit-bridge under valgrind like this:
$ export G_DEBUG=fatal-criticals
$ export G_MESSAGES_DEBUG=cockpit-ws,cockpit-daemon,cockpit-bridge
$ valgrind --trace-children=yes --trace-children-skip='*unix_chkpwd*' \
./cockpit-ws --port 10000 --no-tls
Note that cockpit-session and cockpit-bridge will run from the installed prefix, rather than your build tree.
Some features of Cockpit require a domain to test. Cockpit should work with either Active Directory or IPA.
If you do not have a domain available that you can use, or don't have sufficient privileges on the domain to test Cockpit's features, you can use the IPA server that comes with from Cockpit's integration tests. The domain is called 'cockpit.lan'. On a physical machine, with the cockpit sources checked out, here's how you get it running:
$ cd /path/to/src/cockpit
$ cd ./test
$ ./vm-prep
$ ./vm-download -f ipa
$ ./vm-run -f ipa
The IP address of the IPA server will be printed. The root password
is foobar. The IPA admin password is foobarfoo.
Your client machines (with your web browser) and your server machines (with cockpit running) need to be able to resolve DNS queries against the IPA server. You can do the following to make that happen:
$ sudo -s
# echo -e 'domain cockpit.lan\nnameserver 10.111.111.100\n' > \
/etc/resolv.conf
To test your DNS, the following should succeed without any error messages on both your client machines, and your server with cockpit:
$ host cockpit.lan
Now verify that you can authenticate against the IPA server. See password above.
$ kinit admin@COCKPIT.LAN
Password for admin@COCKPIT.LAN:
BUG: IPA often fails to start up correctly on system boot. You may
have to log into the IPA server and run systemctl start ipa.
ipa bug
Cockpit can perform single sign on authentication via Kerberos. To test and work on this feature, you must have a domain on your network. See section above if you do not.
Use the following guide to configure things, with more troubleshooting advice below:
http://files.cockpit-project.org/guide/sso.html
BUG: The host name of the computer Cockpit is running on should end with the domain name. If it does not, then rename the computer Cockpit is running on: realmd bug
$ sudo hostnamectl set-hostname my-server.domain.com
BUG: If your domain is an IPA domain, then you need to explictly add a service before Cockpit can be used with Single Sign on. The following must be done on the computer running Cockpit. realmd bug
$ sudo -s
# kinit admin@COCKPIT.LAN
# curl -s --negotiate -u : https://f0.cockpit.lan/ipa/json \
--header 'Referer: https://f0.cockpit.lan/ipa' \
--header "Content-Type: application/json" \
--header "Accept: application/json" \
--data '{"params": [["HTTP/my-server.cockpit.lan@COCKPIT.LAN"], {"raw": false, "all": false, "version": "2.101", "force": true, "no_members": false}], "method": "service_add", "id": 0}'
# ipa-getkeytab -q -s f0.cockpit.lan -p HTTP/my-server.cockpit.lan \
-k /etc/krb5.keytab
Now when you go to your cockpit instance you should be able to log in without authenticating. Make sure to use the full hostname that you set above, the one that includes the domain name.
If you want to use Cockpit to connect to a second server. Make sure that second server is joined to a domain, and that you can ssh into it using GSSAPI authentication with the domain user:
$ ssh -o PreferredAuthentications=gssapi-with-mic admin@my-server2.domain.com
If you thought that was nasty and tiresome, it's because it is at present :S
It is possible, although not recommended, to run Cockpit in a privileged container. There is experimental support for this in tools/Dockerfile
To build the privileged container:
$ cd tools/ $ rm -f *.rpm $ ./make-rpms # optional $ docker build .
By running make-rpms first, the latest Cockpit from git HEAD will be built instead of downloading the latest release into the privileged container.