19 KiB
administering and running gitolite
In this document:
please read this first
Unless you know what you're doing, do not do anything manually on the server (except when the documentation says you should, for example to add custom hooks). In particular, adding new repositories or users or changing the access control rules should not be done directly on the server. Things will break. For example, if you manually create a repo on the server, it will not have the required "update" hook, without which there is no access control for pushes.
Most normal (day-to-day) gitolite admin work is done by cloning the gitolite-admin repo from the server to your workstation, making changes to the clone, and pushing those changes back.
The installation steps in the previous section include the steps to do this
clone, so you should already have one on your workstation, in
~/gitolite-admin
. You can of course clone it anywhere else you want and use
that clone.
Either way, make sure you cd
into this clone first.
Note: some of the paths in this document use variable names. Just refer to
~/.gitolite.rc
for the correct values for your installation.
Once you've cloned it, you're ready to add users and repos.
adding users and repos
Do NOT add repos or users directly on the server! You MUST manage the
server by cloning the special 'gitolite-admin' repo on your workstation (git clone git@server:gitolite-admin
), making changes, and pushing them. This
section tells you how to add users and repos.
-
ask each user who will get access to send you a public key. See other sources (for example here) for how to do this
-
rename each public key according to the user's name, with a
.pub
extension, likesitaram.pub
orjohn-smith.pub
. You can also use periods and underscores -
copy all these
*.pub
files tokeydir
in your gitolite-admin repo clone. You can also organise them into various subdirectories ofkeydir
if you wish, since the entire tree is searched. -
edit the config file (
conf/gitolite.conf
in your admin repo clone). See doc/gitolite.conf.mkd in the gitolite source for details on what goes in that file, syntax, etc. Just add new repos as needed, and add new users and give them permissions as required. The users names should be exactly the same as their keyfile names, but without the.pub
extension -
when done, commit your changes and push. Any new repos you specified will automatically be created (empty, but clonable) and users' access will be updated as needed.
using hooks
custom hooks
You can supply your own, custom, hook scripts if you wish. Install gitolite as usual, then:
- go to ~/.gitolite/hooks/common on the server and put your new hook there
- now run "gl-setup" again
You can use this procedure to install new hooks as well as to update hooks that you had previously installed.
IMPORTANT WARNINGS
-
The
update
hook inhooks/common
is what implements all the branch-level permissions in gitolite. If you fiddle with the hooks directory, please make sure you do not mess with this file accidentally, or all your fancy per-branch permissions will stop working. -
Do not under any conditions put anything in
hooks/gitolite-admin
-- nothing in gitolite requires you to do anything here. Leave it alone!
"gl-post-init" hook
Sometimes it is necessary to do something whenever a new repo is created. If you need this functionality, just supply a hook called "gl-post-init" with whatever code you want in it.
"gl-pre-git" hook
Although git has lots of nice hooks you can tap into, they all run only on a push. There's nothing that runs on a fetch or a clone, and there's no way to run something before git-receive-pack or git-upload-pack, (as the case may be) are invoked.
That's what the gl-pre-git
hook is for. If an executable hook called
gl-pre-git
is present, it will be invoked with the current directory set to
repo.git
, and with a single argument which will be either R
or W
depending on what the client is trying to do. The environment variables
GL_USER
and GL_REPO
are available. STDOUT will be forced to STDERR before
it is called, to avoid confusing the client.
If the code returns anything other than 0, gitolite will terminate the
operation (i.e., not run git at all), just like many git hooks do, so make
sure you end with exit 0
or equivalent.
hook chaining
Sometimes you need to use git hooks for your own purposes (site-local validations, CI integration, email notifications, or the ever popular "live website update"!). However, the hooks you want to use may already be in use by gitolite.
This section will tell you what to do in such cases. First, let's list the hooks that gitolite uses:
-
The
update
hook is used in all repos and is critical to gitolite's access control! -
The
post-receive
hook is used in all repos but only if mirroring has been enabled. Shipped aspost-receive.mirrorpush
, it is renamed to 'post-receive' and installed as part of the mirroring setup. -
The
post-update
hook is used in thegitolite-admin
repo only, to "compile" the configuration and so on.
To run your own 'update' hook, just put it in a file called update.secondary
and install it as a hook. Gitolite's update hook will automatically chain to
it, taking care to pass it the same 3 arguments the original update hook
received from git.
In addition, gitolite now contains the basic infrastructure to support multiple 'update' hooks without having to remember to chain them yourself. See
hooks/common/update.secondary.sample
for instructions.
For post-receive
, (if using mirroring) do the opposite. You're normally
expected to rename the shipped 'post-receive.mirrorpush' to 'post-receive',
but don't do this. Instead, simply run hooks/post-receive.mirrorpush
at the
end of your hook code. Do not worry about replicating STDIN (the documented
way in which a post-receive hook receives its input) because the mirroring
code does not use it.
To run your own post-update
hook on normal repos, just install a hook called
'post-update' the usual way. It'll be installed on all normal repos but not
on the special gitolite-admin repo. If you need that for the gitolite-admin
repo, you'll have to call it post-update.secondary
.
Finally, these names ('update.secondary' and 'post-update.secondary') are merely the defaults. You can change them to anything you want; look in conf/example.gitolite.rc for details.
environment variables available to hooks
The following environment variables are set, and may be useful for any custom processing you wish to do in your hook code:
GL_USER
-- the user doing the pushGL_REPO
-- the reponameGL_REPO_BASE_ABS
-- the absolute base path where all the repos are kept
The following variables are also set, but are generally less useful:
GL_BINDIR
-- where all the binaries liveGL_ADMINDIR
-- common directory for many gitolite things
other features
moving pre-existing repos into gitolite
It's best to split this into different use cases.
Case 1 -- few repos: This is for moving one or two repos at a time, when you have a copy of the repo on your workstation. It is also the only way if you have push rights to the admin repo but no shell privileges on the server.
-
let gitolite create it as a brand new repo as described in the section on "adding users and repos" at the top
-
cd to the clone on your workstation. Make sure all the branches are correct and no extra stuff, "temp" branches, etc., are present
-
now run these two commands
git push --all git@server:reponame git push --tags git@server:reponame
-
(You could also use "git push --mirror" instead of separately doing branches and tags, but that will carry across your remote refs also, and typically you may not want that. Anyway please do a
git ls-remote git@server:repo
to make sure all the stuff you want went through, and is named correctly).
Case 2 -- many repos: This is when you have many existing repos to add, and they're all bare (as good little server repos should be) and you have shell access on the server. Here's how to do it; please note the order is important here:
-
make doubly sure they're bare repos ;-)
-
log on to the server and copy the repos to
$REPO_BASE
(which defaults to~/repositories
), making sure that the directory names end in ".git". -
back on your workstation, add each repo (without the
.git
suffix) toconf/gitolite.conf
in your gitolite-admin repo clone. Give some user (even a non-existent one like "DUMMY" is fine) at least "R" access to these repos. Then add, commit, push.
Case 3 -- far too many repos (or your initials are JH ;-): This is when you're like Case 2, except you have so many repos that step 3 becomes too cumbersome (even with a script doing it for you).
Assuming you can group your repo names into various patterns, and can use similar access control lines within each such group, you can use gitolite's "wildcard repos" feature.
First read doc/wildcard-repositories.mkd, or at least skim through it, to understand the basic concept. Then do this:
-
do step 1 just like step 1 in Case 2 above
-
ditto for step 2
-
for each repo, determine who the owner should be and create files called
gl-creater
(note spelling!) in each repo. The file should contain exactly one line with the owner name. -
run
gl-setup
again (you don't need to supply a pub key filename) -
finally add the repos to the conf, maybe something like this, (in this example, the owner name was the second component of the repo path), and add/commit/push:
repo pub/CREATOR/..* C = @developers RW+ = CREATOR RW = WRITERS R = READERS
Details
-
why is the order of steps different in case 1 and case 2?
Because in case 2, the actual data is coming from an OS 'cp' (copy) command, not via a normal push like in case 1. Since that happens outside gitolite, it's easier to do it first, then tell gitolite about the repo so it can add hooks. (If you tell gitolite first, it will create an empty repo as soon as you push, then your 'cp' will have to overwrite those files, but you'll then lose gitolite's hooks, etc. A bit more messy).
-
what's with the
gl-creater
file in case 3?What doc/wildcard-repositories.mkd does not explain is how ownership is recorded in gitolite: the
gl-creater
file contains the owner name. If you want to "pretend" these repos were created by some user, you need to add that in. That user then gets whatever access you gave to "CREATOR" in the access rules (in our example, that wasRW+
). -
why does case 3 need the
gl-setup
command?An admin push only checks hooks on normal (non-wildcard) repos. It would be too timetaking otherwise. Running
gl-setup
forces it to do this more aggressively than an admin push, looking at wildcard repos as well as normal ones.
In the end, it all boils down to (a) making sure the update
hook is correct
on all repos, wild or normal, and (b) making sure gl-creater
contains the
owner name for wild repos. The rest of the setup is in the conf file.
moving the whole thing from one server to another
[NOTE: I would appreciate help testing these instructions]
Just copying everything won't work unless everything on the new server is exactly the same. I suggest you don't try it unless you know what you're doing.
Assumptions
- you have not changed
$REPO_BASE
on either of the servers; if you did, substitute accordingly - the admin's name is "YourName" -- again, substitute accordingly!
- the "hosting user" on both servers is "git". Substitute whatever you're actually using (for example, if you're installing using RPM/DEB, this would be "gitolite")
There are many ways of doing this, but the most generic set of steps are given below. Please follow all the steps; do not skip or improvise! Ask me if things are not clear -- you can help me fine tune this document :-)
-
(old server) disable the old server so your users will not push any changes to it. There are several ways to do this, but the simplest is to insert this line at the top of
~/.gitolite.rc
on the old server:exit 1;
-
(new server) copy the repos to the new server, except the
gitolite-admin
repo and files calledgitolite-hooked
in thehooks
directory of each repo.That sounds complicated but it's not. It's just:
cd $HOME rsync -a olduser@oldhost:repositories . mv repositories/gitolite-admin.git $HOME/old-gitolite-admin.git find repositories -name gitolite-hooked | xargs rm
-
(workstation, new server) install gitolite normally on your new server. Use whatever install method suits you, but you must use the same name for the admin ("YourName" in the install instructions). You may use a different keypair if you need to, or use the same one that currently gets access to the old server.
-
(new server) edit the
~/.gitolite.rc
file to match the settings on the old server, if needed. Do not copy the entire file outright -- some of the variables (notablyGL_PACKAGE_CONF
andGL_PACKAGE_HOOKS
) are installation dependent and should not be touched! Do a diff or a vimdiff and copy across only what you know you changed on the old server. -
(workstation) push the config to the new server. To do this, go to your admin clone, and:
-
if you used a different keypair when installing to the new server, copy that pubkey to this clone into
keydir/Yourname.pub
, then add and commit the change to the pubkeycd gitolite-admin cp path/to/new/YourName.pub keydir/YourName.pub git add keydir git commit -m "new server, new key"
-
if you did not use a different keypair, just make a dummy commit
git commit -m "new server" --allow-empty
-
set the URL for the new server
git remote --set-url origin git@newserver:gitolite-admin
-
push the config, including past history
git push -f
-
And that should be that!
specifying gitweb and daemon access
This is a feature that I personally do not use (corporate environments don't like unauthenticated access of any kind to any repo!), but someone wanted it, so here goes.
Gitolite has two pre-defined, "special", usernames: daemon
and gitweb
.
To make a repo or repo group accessible via "git daemon", just give read
permission to the special user "daemon". Similarly, give read permission to
gitweb
to allow the gitweb CGI to show the repo. Something like this:
repo foo bar baz
R = gitweb daemon
This gives you a quick way to offer multiple repos up for gitweb and/or daemon access.
However, setting a description for the project also enables gitweb permissions so you can do it that way if you want. Of course in this case you have to deal with each repo separately. Add lines like this to gitolite.conf:
foo = "some description"
bar = "some other description"
baz = "yet another description"
You can also specify an owner for gitweb to show, if you like; for example I might use:
gitolite "Sitaram Chamarty" = "fast, secure, fine-grained, access control for git"
Note that gitolite does not install or configure gitweb/git-daemon -- that is a one-time setup you must do separately. All gitolite does is:
- for daemon, create the file
git-daemon-export-ok
in the repository - for gitweb, add the repo (plus owner name, if given) to the list of
projects to be served by gitweb (see the config file variable
$PROJECTS_LIST
, which should have the same value you specified for$projects_list
when setting up gitweb) - put the description, if given, in
$repo/description
The "compile" script will keep these files consistent with the config settings -- this includes removing such settings/files if you remove "read" permissions for the special usernames or remove the description line.
Please note that giving permissions to these special users via @all
(that is, using either repo @all
or R = @all
), will not work unless you
set the rc-file variable $GL_ALL_INCLUDES_SPECIAL
to 1
. Also, NOTE
that giving them read access to repo @all
means the gitolite-admin
repo is
also accessible. It is upto you to decide if that is OK in your
environment.
custom git config
The custom hooks feature is a blunt instrument -- all repos get the hook you specified and will run it. In order to make it a little more fine-grained, you could set your hooks to only work if a certain "gitconfig" variable was set. Which means we now need a way to specify "git config" settings on a per repository basis.
Thanks to Teemu (teemu dot matilainen at iki dot fi), gitolite now does this very easily. For security reasons, this can only be done from the master config file (i.e., if you're using delegation, the delegated admins cannot specify git config settings).
Please see doc/gitolite.conf.mkd
for syntax and limitations. Also note that
this feature is disabled by default. Read the comments on a variable called
GL_GITCONFIG_KEYS
in the rc file documentation, then set it to some
appropriate value, to enable this feature.