8 ms·

Self-hosted Git Server

115 points by emilyst 13y ago

taylorlapeyre 13y ago
An elegant solution for you, no doubt. However, I couldn't part with the excellent tools that GitHub has developed (and will continue to develop) over the years.
- michaelmior 13y ago
  I still use GitHub quite a lot, but I've started using a private installation of GitLab[1] for private projects and I quite love it. So far, I've only been using it for myself, so I'll admit I haven't really touched the collaborative features, but it does seem to have a lot of the nice features from GitHub. https://www.gitlab.com/ https://www.gitlab.com/
  - zrail 13y ago
    Yeah, GitLab is great. I wanted a quick solution without many dependencies that would also act as a portfolio, so I went with a custom thing. GitLab is definitely something to consider.
    - neumann 13y ago
      I use GitLab at work, but for home, the amazing GitBucket does the trick and took a total of 30 seconds to install https://github.com/takezoe/gitbucket https://github.com/takezoe/gitbucket
      - sytse 13y ago
        If you want to setup GitLab quickly please check out the packages https://www.gitlab.com/downloads/ https://www.gitlab.com/downloads/ (I'm a co-founder of GitLab.com)
  - HarshaThota 13y ago
    For those of you on Windows, Bonobo[1] is a good alternative. [1] http://bonobogitserver.com/ http://bonobogitserver.com/
    - voltagex_ 13y ago
      I wonder how much effort it'd take to get git+ssh:// transport working with that server.
zrail 13y ago
Sorry about the occasional 502s. I added a smattering of caching to the project pages so it should be better now.
- stefan_kendall3 13y ago
  This is my point exactly. Self hosting is maintenance. In the time you got 502s, my github pages stayed up. It is grossly inefficient to host your own source control. Libraries and third party services exist for a reason.
  - zrail 13y ago
    Life is built of tradeoffs and compromises. For you, github pages makes total sense. For me, not so much. Saying that my little VPS is "grossly inefficient" compared to your GitHub repositories is like comparing apples to... I dunno, like a rock? or a monkey? Something else entirely, that's for sure.
    - camus2 13y ago
      It only proves you did not build your solution with scalability in mind. Once you get more traffic and the hosting bill goes up, i'm pretty sure you'll reconsider self hosting for the sake of self hosting.
      - zrail 13y ago
        The page in the original post has received 16,000 pageviews in the last 12 hours. Admittedly the project pages were not performing very well initially but after I added caching they haven't even blinked, and the dual core VPS (which runs a dozen other services besides this) is ticking along at a load average of 0.63. Tell me again how I didn't design for scalability. I'd also love to hear your definition of "more traffic".
jmoiron 13y ago
I spent years out in the woods with my own projects page that nobody could find and very few cared about, happily cutting trees down with nary a witness. I wanted control over backups, presentation, availability. I was terrified of relying on services that could disappear. I found self promotion distasteful, and I was happy to do the work just for its own sake, for the enjoyment of that process. But then I wrote something people actually used. And they wanted it on Bitbucket, and then on Github, so they could contribute to it and track its progress. Now that I've come in from my Zen training out in the wilderness, I feel that the old me who wanted "control" over all of my creations was immature and selfish, abstaining from participation in a community and helping no one but myself. I think that the social aspect of hosted coding sites, both for collaboration and exposure, is much more valuable than the control you get from running the whole show yourself.
- zrail 13y ago
  Honestly, I'm not super concerned about GitHub disappearing in a puff of octocat ink one day. I know they're around for the long haul and I think they've done great things for open source. No, I'm more concerned about branding. If someone comes to my site and I direct them somewhere else, I lose them. If they stay on my site and browse around, my site's look and feel is indelibly linked with my code, my projects, and ultimately me as a professional software developer.
  - cnaut 13y ago
    Is branding the biggest reason for hosting your own git server?
    - zrail 13y ago
      For me personally, I guess the biggest reason is being able to control the user's experience, as well as my own. Plus I think it's neat, and it was a fun project to do on a Sunday afternoon :)
      - comex 13y ago
Nux 13y ago
Also check https://github.com/takezoe/gitbucket https://github.com/takezoe/gitbucket Installation is as easy as downloading it and running: java -jar gitbucket.war
- stefan_kendall3 13y ago
  Disingenuous. What happens when you need to restart your server because of some extreme necessary security patch? How do you manage backups? Since it's running on java, you probably don't want to expose this server to the internet unless you actively want to maintain updates to JDK versions. So now you're on the hook for that. Self-hosting is never "simple".
  - maxerickson 13y ago
    I do not believe the GP was trying to be misleading. That's what calling it disingenuous means. You think it is a bad idea or something.
    - asdafa 13y ago
      Problem there might be that 'disingenuous' can be a false friend for non-english native speakers... But I do tend to agree with his comment. Self hosting is never simple, it's one more thing you have to take care of. Yes there might be cases where self hosting something is a requirement but, personally, for all the things that are strictly not work related and things that "just have to work" (e.g.: email, im, code hosting for free time projects) I use third party services, unfortunately I don't have enough time to take care of those things, and since I have to take care of similar things at work, at least for my own spare time I want something that helps me focus on what I really want to do.
  - sdesol 13y ago
    "Since it's running on java, you probably don't want to expose this server to the internet" Is latest versions from tomcat/jetty/etc. really known for having major security holes? More so than apache/nginx/etc. Maybe you are confusing the recent Java applets security issue with Java in general. Java has got to be one of the most well funded and developed technology out there, due to peoples reliance to it in enterprise. People also don't normally run their Java web server as root which adds a bit more security. If there is something about Java security that makes you so worried, I would love to hear about it. As it will probably be news to me.
lholden 13y ago
The fantastic thing about distributed version control systems like GIT is that you never actually have to depend on a single repository. Use Github, Bitbucket, and your own "server" all at once. None have to be read-only. Push and pull from multiple sources. Decide that Github is no longer the place for you? No problem! People contributing to your projects should already know where to push to reach everywhere it needs to go and they only need to stop pushing to github. No reason to rely on a "central" repo.
- _frog 13y ago
  Worth noting though, a lot of the value of GitHub in particular, and to a lesser extent many other services, are extra features like issue trackers, wikis and so on that act to lock you in.
  - lholden 13y ago
    I can certainly see services like wiki and issue tracker hosting being big value-adds with GitHub. Thankfully GitHub does allow you to export your data. (And to be honest, I feel it would be a mistake to use these services if you couldn't export your data.) The wiki system on GitHub is based out of it's own git repo (https://github.com/[user]/[repo].wiki.git https://github.com/[user]/[repo].wiki.git). If you could be bothered to pull/push when changes are made, I'd imagine this content could then be made distributed. While not distributed/decentralized... GitHub does lets you export your existing issues. The resulting archive can then be imported into services like bitbucket, pivotal, and etc with a little work. There are solutions for distributed/decentralized issue tracking, but I haven't seen any great way for users to interact with them. (SD and Bugs Everywhere, other users have posted links in their replies.)
geerlingguy 13y ago
Here's another example, using Ansible and a few simple roles from Ansible Galaxy: https://github.com/geerlingguy/ansible-vagrant-examples/tree/master/gitlab https://github.com/geerlingguy/ansible-vagrant-examples/tree... Give Vagrant + Ansible about 15 minutes, and you'll have a nice little local GitLab server running. Modify the Ansible playbook slightly, to point it at a VM on Digital Ocean or elsewhere, and you'll have a nice little hosted copy of something like GitHub. I run an instance of GitLab on a Digital Ocean droplet, and back it up to a cheap RamNode VPS; all my GitHub repos are pulled daily to repos on the GitLab server (for backup purposes—I use the GitHub repos primarily), and I also use this GitLab server as a central repo to which I push changes for my sites/services, and pull from on production servers. There are many good reasons to put certain repositories on GitHub, and there are some good reasons to host some repositories yourself (or to just have one local working copy, with a backup).
- HeyImAlex 13y ago
  I've been playing with this same setup for the last two weeks at work. I've got our full dev server config in ansible and stored in a repo that has a post receive hook that self-runs the playbook, and an instance of git lab running in a container and routed via nginx. Gitolite is pretty robust, but after setting it up I kept on thinking of features I would like or need and I realized I was just rewriting git lab :) there are some docker images that make setting it up cake, and a full installer available for centos and Ubuntu. Also I think I've used some of your roles off galaxy :) thanks for writing them haha
- showerst 13y ago
  I prefer ansible too, but FWIW DigitalOcean has a pre-configured gitlab droplet as well. You can literally have it live in < 5 minutes, including registering and configuring a domain.
616c 13y ago
I checked out this guy's site. I understand he is reinventing the wheel. I respect that and do not mean anything negative about that. A lot of us are here to understand the wheel and build for ourselves. There is good value for that. What I do not understand, with all due respect, is the value of Github, and to a lesser extent Bitbucket, is in the features outside of the VCS core and how most people realize there is nothing Git-like to replace the actual project management tools where people find Github value. That is, bug tracking/issue tracking is the killer feature. To self-host this is problematic. This is not to say there are not good solutions. Traditionally that is Bugzilla, Trac, Redmine, and more recently a la ArchLinux and Music Player Daemon (MPD) I have seen Mantis BT. It looks interesting. The problem is, as others point out, maintenance, and adding the centralized to the decentralized, thus the point of Git, Mercurial, and others. I noticed this guy hosts his own repos (I tried links to bugsplat.rb) and it did not load, but I assume there is not bug tracking, despite the name. Even this year, I decided to look into the state of DVCS integrated bug tracking. Very few tools exist, or many have problems. BE, bugseverywhere, kind of exists and has seen contributions as of last year. Ticgit was forked to ticgit-ng, but ironically that is a Github project and its bugs are in the Github issue tracker. There was a very cool Perl project, which seems abandoned, called SimpleDefects (SD), which also wanted to do distributed bug tracking. It was going to sync with Github and other issue tracker systems with decentralizaton, so you could git pull for bugs as well. This one has not seen updates for years, as I can tell, but might be the coolest of them. Unfortunately, keeping you whole project decentralized is difficult. I have begun to look into fossil again, despite what people here say, because it might be ugly, but no VCS has its own bug tracker or wiki integration, all written in C. It is the SCM and bug tracker for Sqlite, Tcl, and even for all of NetBSD packages. The last really surprised me. I think for the little guy, that is ideal. Monotone might be worth revisiting (I had multiple problems 3 or 4 months ago because it was embedding Lua and they had not come up to the 5.1-5.2 API changes and builds failed; everyone is hard to find), even with the only data about it is there page and links to snarky Linus Torvalds jokes about the horror of OO data structures and modern C++ programming for a SCM. In short, you should know to host your own Git repos (they are designed, with the native package tools in Git or others) to host simply on a website. It is best for even the lowest footprint web servers, shared hosting or not. What git does not have, is the beautiful features that keep people on Github all day. I really wish distributed bug tracking advanced, but no one is interested. This topic only comes up when Github is down (or maybe in this case but Github is having publicity problems today). If people made a good tool like SD, that has its own bug tracking that syncs well with Github or (insert hosted SCM here), that would be fantastic. I could be more relaxed, principally, by relying on such tools. Back to read the Fossil manual.
hipaulshi 13y ago
Well. I have to be honest. I used to do just that. And it is not fun to worry about backups and server breaking down. Outsourcing this part to Github will probably save a lot of developer/IT headache if you are a small company.
mixologic 13y ago
What about all the problems that are inherent to hosting anything ? It's not hard to set up a server that will respond to a request and deliver data. It's hard to make it do that when the power goes out, when you anger somebody and they decide to ddos you, or when your code/service/whatever becomes insanely popular and whatever mac mini you run it on is overheating. It's hard to keep up with all the security patches required to keep your box from getting pwned. It's hard to negotiate the fine print with all the TOS's you've signed with your internet provider.
- zrail 13y ago
  I'm sorry, I should have clarified. My private repos are running on a VM in the living room, but the public-facing website and repos are all on a Digital Ocean VPS.
sqrt17 13y ago
For private repositories, I've discovered a software called SCM Server. It lets you add different users, with passwords, and create SVN/Hg/Git repositories with each their own access list.
aleprok 13y ago
I have registered in January 27 2009 to github, but I have been running my own git repository on a dedicated server for quite some time. I have never really been active at github or any other third party repository service. The reason I want to self manage my hosting is that I can get cheaper private repositories.
tenfingers 13y ago
I also use my private repositories (simply setup using gitolite), but I simply also push to GitHub for the added discoverability. I tend not to like GitHub however. The wiki/issue tracker actually require you to be online, which is the best thing about git: being able to work without network connection. I'd like to find some project that syncs github issues with sd (http://syncwith.us/sd/ http://syncwith.us/sd/) or bugseverywhere (http://www.bugseverywhere.org/ http://www.bugseverywhere.org/), so that I could actually work on the issues without being online (the bigger pain point are attachments which need to be downloaded). Another thing which I consider inferior to patches by mail are pull requests. Honestly, it's very rare that I can accept a patch without some reformatting or minor corrections. I also never ask people to do such minor reformatting just to be able to click 'accept'. Again, pull requests cannot be worked-on or accepted offline. Similarly, my patches sent to other projects are similarly handled by other authors.
- beagle3 13y ago
  Fwiw, pull requests are merely branches im the repository (named pr-* if i'm not mistaken)
rsgalloway 13y ago
heh, what at the odds: two libs that give you object-oriented access to git repos both coincidentally named the exact same thing: python: github.com/rsgalloway/grit ruby: github.com/mojombo/grit
szhu 13y ago
It looks there are two main reasons to have a private Git server: to archive your own code, and to have control over how visitors and collaborators view your shared code. If you want a place to store or archive personal projects (because you don't want to pay for private Git hosting?) features like wikis and issues become rather irrelevant, and if you want to move that data to or from GitHub, that takes some effort to be explicitly exported instead of just adding a new remote. If you're trying to keep your own code, running your own Git server is simple if you have SSH (or not even SSH if you want to create a local "server"). The Git website has a simple tutorial for this: http://git-scm.com/book/en/Git-on-the-Server-Setting-Up-the-Server http://git-scm.com/book/en/Git-on-the-Server-Setting-Up-the-... I've also simplified this workflow into a tiny command, "mygit": https://github.com/interestinglythere/mygit https://github.com/interestinglythere/mygit As for running a GitHub clone, unless you have some really edge-case needs or you really don't want to pay for private collaborative repos, your time is probably better spent maintaining your actual code. In addition, having your code on a social site will make it more receptive to contributors. If you don't want contributors, why do you need a /Git/ server? (If it's because that's how you like to manage your code, see my first point.)

115 points by emilyst 13y ago

An elegant solution for you, no doubt. However, I couldn't part with the excellent tools that GitHub has developed (and will continue to develop) over the years.

michaelmior 13y ago
I still use GitHub quite a lot, but I've started using a private installation of GitLab[1] for private projects and I quite love it. So far, I've only been using it for myself, so I'll admit I haven't really touched the collaborative features, but it does seem to have a lot of the nice features from GitHub. https://www.gitlab.com/ https://www.gitlab.com/
- zrail 13y ago
  Yeah, GitLab is great. I wanted a quick solution without many dependencies that would also act as a portfolio, so I went with a custom thing. GitLab is definitely something to consider.
  - neumann 13y ago
    I use GitLab at work, but for home, the amazing GitBucket does the trick and took a total of 30 seconds to install https://github.com/takezoe/gitbucket https://github.com/takezoe/gitbucket
    - sytse 13y ago
      If you want to setup GitLab quickly please check out the packages https://www.gitlab.com/downloads/ https://www.gitlab.com/downloads/ (I'm a co-founder of GitLab.com)
- HarshaThota 13y ago
  For those of you on Windows, Bonobo[1] is a good alternative. [1] http://bonobogitserver.com/ http://bonobogitserver.com/
  - voltagex_ 13y ago
    I wonder how much effort it'd take to get git+ssh:// transport working with that server.

zrail 13y ago

Sorry about the occasional 502s. I added a smattering of caching to the project pages so it should be better now.

stefan_kendall3 13y ago
This is my point exactly. Self hosting is maintenance. In the time you got 502s, my github pages stayed up. It is grossly inefficient to host your own source control. Libraries and third party services exist for a reason.
- zrail 13y ago
  Life is built of tradeoffs and compromises. For you, github pages makes total sense. For me, not so much. Saying that my little VPS is "grossly inefficient" compared to your GitHub repositories is like comparing apples to... I dunno, like a rock? or a monkey? Something else entirely, that's for sure.
  - camus2 13y ago
    It only proves you did not build your solution with scalability in mind. Once you get more traffic and the hosting bill goes up, i'm pretty sure you'll reconsider self hosting for the sake of self hosting.
    - zrail 13y ago
      The page in the original post has received 16,000 pageviews in the last 12 hours. Admittedly the project pages were not performing very well initially but after I added caching they haven't even blinked, and the dual core VPS (which runs a dozen other services besides this) is ticking along at a load average of 0.63. Tell me again how I didn't design for scalability. I'd also love to hear your definition of "more traffic".

jmoiron 13y ago

I spent years out in the woods with my own projects page that nobody could find and very few cared about, happily cutting trees down with nary a witness. I wanted control over backups, presentation, availability. I was terrified of relying on services that could disappear. I found self promotion distasteful, and I was happy to do the work just for its own sake, for the enjoyment of that process. But then I wrote something people actually used. And they wanted it on Bitbucket, and then on Github, so they could contribute to it and track its progress. Now that I've come in from my Zen training out in the wilderness, I feel that the old me who wanted "control" over all of my creations was immature and selfish, abstaining from participation in a community and helping no one but myself. I think that the social aspect of hosted coding sites, both for collaboration and exposure, is much more valuable than the control you get from running the whole show yourself.

zrail 13y ago
Honestly, I'm not super concerned about GitHub disappearing in a puff of octocat ink one day. I know they're around for the long haul and I think they've done great things for open source. No, I'm more concerned about branding. If someone comes to my site and I direct them somewhere else, I lose them. If they stay on my site and browse around, my site's look and feel is indelibly linked with my code, my projects, and ultimately me as a professional software developer.
- cnaut 13y ago
  Is branding the biggest reason for hosting your own git server?
  - zrail 13y ago
    For me personally, I guess the biggest reason is being able to control the user's experience, as well as my own. Plus I think it's neat, and it was a fun project to do on a Sunday afternoon :)
    - comex 13y ago

Nux 13y ago

Also check https://github.com/takezoe/gitbucket https://github.com/takezoe/gitbucket Installation is as easy as downloading it and running: java -jar gitbucket.war

stefan_kendall3 13y ago
Disingenuous. What happens when you need to restart your server because of some extreme necessary security patch? How do you manage backups? Since it's running on java, you probably don't want to expose this server to the internet unless you actively want to maintain updates to JDK versions. So now you're on the hook for that. Self-hosting is never "simple".
- maxerickson 13y ago
  I do not believe the GP was trying to be misleading. That's what calling it disingenuous means. You think it is a bad idea or something.
  - asdafa 13y ago
    Problem there might be that 'disingenuous' can be a false friend for non-english native speakers... But I do tend to agree with his comment. Self hosting is never simple, it's one more thing you have to take care of. Yes there might be cases where self hosting something is a requirement but, personally, for all the things that are strictly not work related and things that "just have to work" (e.g.: email, im, code hosting for free time projects) I use third party services, unfortunately I don't have enough time to take care of those things, and since I have to take care of similar things at work, at least for my own spare time I want something that helps me focus on what I really want to do.
- sdesol 13y ago
  "Since it's running on java, you probably don't want to expose this server to the internet" Is latest versions from tomcat/jetty/etc. really known for having major security holes? More so than apache/nginx/etc. Maybe you are confusing the recent Java applets security issue with Java in general. Java has got to be one of the most well funded and developed technology out there, due to peoples reliance to it in enterprise. People also don't normally run their Java web server as root which adds a bit more security. If there is something about Java security that makes you so worried, I would love to hear about it. As it will probably be news to me.

lholden 13y ago

The fantastic thing about distributed version control systems like GIT is that you never actually have to depend on a single repository. Use Github, Bitbucket, and your own "server" all at once. None have to be read-only. Push and pull from multiple sources. Decide that Github is no longer the place for you? No problem! People contributing to your projects should already know where to push to reach everywhere it needs to go and they only need to stop pushing to github. No reason to rely on a "central" repo.

_frog 13y ago
Worth noting though, a lot of the value of GitHub in particular, and to a lesser extent many other services, are extra features like issue trackers, wikis and so on that act to lock you in.
- lholden 13y ago
  I can certainly see services like wiki and issue tracker hosting being big value-adds with GitHub. Thankfully GitHub does allow you to export your data. (And to be honest, I feel it would be a mistake to use these services if you couldn't export your data.) The wiki system on GitHub is based out of it's own git repo (https://github.com/[user]/[repo].wiki.git https://github.com/[user]/[repo].wiki.git). If you could be bothered to pull/push when changes are made, I'd imagine this content could then be made distributed. While not distributed/decentralized... GitHub does lets you export your existing issues. The resulting archive can then be imported into services like bitbucket, pivotal, and etc with a little work. There are solutions for distributed/decentralized issue tracking, but I haven't seen any great way for users to interact with them. (SD and Bugs Everywhere, other users have posted links in their replies.)

geerlingguy 13y ago

Here's another example, using Ansible and a few simple roles from Ansible Galaxy: https://github.com/geerlingguy/ansible-vagrant-examples/tree/master/gitlab https://github.com/geerlingguy/ansible-vagrant-examples/tree... Give Vagrant + Ansible about 15 minutes, and you'll have a nice little local GitLab server running. Modify the Ansible playbook slightly, to point it at a VM on Digital Ocean or elsewhere, and you'll have a nice little hosted copy of something like GitHub. I run an instance of GitLab on a Digital Ocean droplet, and back it up to a cheap RamNode VPS; all my GitHub repos are pulled daily to repos on the GitLab server (for backup purposes—I use the GitHub repos primarily), and I also use this GitLab server as a central repo to which I push changes for my sites/services, and pull from on production servers. There are many good reasons to put certain repositories on GitHub, and there are some good reasons to host some repositories yourself (or to just have one local working copy, with a backup).

HeyImAlex 13y ago
I've been playing with this same setup for the last two weeks at work. I've got our full dev server config in ansible and stored in a repo that has a post receive hook that self-runs the playbook, and an instance of git lab running in a container and routed via nginx. Gitolite is pretty robust, but after setting it up I kept on thinking of features I would like or need and I realized I was just rewriting git lab :) there are some docker images that make setting it up cake, and a full installer available for centos and Ubuntu. Also I think I've used some of your roles off galaxy :) thanks for writing them haha
showerst 13y ago
I prefer ansible too, but FWIW DigitalOcean has a pre-configured gitlab droplet as well. You can literally have it live in < 5 minutes, including registering and configuring a domain.

616c 13y ago

I checked out this guy's site. I understand he is reinventing the wheel. I respect that and do not mean anything negative about that. A lot of us are here to understand the wheel and build for ourselves. There is good value for that. What I do not understand, with all due respect, is the value of Github, and to a lesser extent Bitbucket, is in the features outside of the VCS core and how most people realize there is nothing Git-like to replace the actual project management tools where people find Github value. That is, bug tracking/issue tracking is the killer feature. To self-host this is problematic. This is not to say there are not good solutions. Traditionally that is Bugzilla, Trac, Redmine, and more recently a la ArchLinux and Music Player Daemon (MPD) I have seen Mantis BT. It looks interesting. The problem is, as others point out, maintenance, and adding the centralized to the decentralized, thus the point of Git, Mercurial, and others. I noticed this guy hosts his own repos (I tried links to bugsplat.rb) and it did not load, but I assume there is not bug tracking, despite the name. Even this year, I decided to look into the state of DVCS integrated bug tracking. Very few tools exist, or many have problems. BE, bugseverywhere, kind of exists and has seen contributions as of last year. Ticgit was forked to ticgit-ng, but ironically that is a Github project and its bugs are in the Github issue tracker. There was a very cool Perl project, which seems abandoned, called SimpleDefects (SD), which also wanted to do distributed bug tracking. It was going to sync with Github and other issue tracker systems with decentralizaton, so you could git pull for bugs as well. This one has not seen updates for years, as I can tell, but might be the coolest of them. Unfortunately, keeping you whole project decentralized is difficult. I have begun to look into fossil again, despite what people here say, because it might be ugly, but no VCS has its own bug tracker or wiki integration, all written in C. It is the SCM and bug tracker for Sqlite, Tcl, and even for all of NetBSD packages. The last really surprised me. I think for the little guy, that is ideal. Monotone might be worth revisiting (I had multiple problems 3 or 4 months ago because it was embedding Lua and they had not come up to the 5.1-5.2 API changes and builds failed; everyone is hard to find), even with the only data about it is there page and links to snarky Linus Torvalds jokes about the horror of OO data structures and modern C++ programming for a SCM. In short, you should know to host your own Git repos (they are designed, with the native package tools in Git or others) to host simply on a website. It is best for even the lowest footprint web servers, shared hosting or not. What git does not have, is the beautiful features that keep people on Github all day. I really wish distributed bug tracking advanced, but no one is interested. This topic only comes up when Github is down (or maybe in this case but Github is having publicity problems today). If people made a good tool like SD, that has its own bug tracking that syncs well with Github or (insert hosted SCM here), that would be fantastic. I could be more relaxed, principally, by relying on such tools. Back to read the Fossil manual.

hipaulshi 13y ago

Well. I have to be honest. I used to do just that. And it is not fun to worry about backups and server breaking down. Outsourcing this part to Github will probably save a lot of developer/IT headache if you are a small company.

mixologic 13y ago

What about all the problems that are inherent to hosting anything ? It's not hard to set up a server that will respond to a request and deliver data. It's hard to make it do that when the power goes out, when you anger somebody and they decide to ddos you, or when your code/service/whatever becomes insanely popular and whatever mac mini you run it on is overheating. It's hard to keep up with all the security patches required to keep your box from getting pwned. It's hard to negotiate the fine print with all the TOS's you've signed with your internet provider.

zrail 13y ago
I'm sorry, I should have clarified. My private repos are running on a VM in the living room, but the public-facing website and repos are all on a Digital Ocean VPS.

sqrt17 13y ago

For private repositories, I've discovered a software called SCM Server. It lets you add different users, with passwords, and create SVN/Hg/Git repositories with each their own access list.

aleprok 13y ago

I have registered in January 27 2009 to github, but I have been running my own git repository on a dedicated server for quite some time. I have never really been active at github or any other third party repository service. The reason I want to self manage my hosting is that I can get cheaper private repositories.

tenfingers 13y ago

I also use my private repositories (simply setup using gitolite), but I simply also push to GitHub for the added discoverability. I tend not to like GitHub however. The wiki/issue tracker actually require you to be online, which is the best thing about git: being able to work without network connection. I'd like to find some project that syncs github issues with sd (http://syncwith.us/sd/ http://syncwith.us/sd/) or bugseverywhere (http://www.bugseverywhere.org/ http://www.bugseverywhere.org/), so that I could actually work on the issues without being online (the bigger pain point are attachments which need to be downloaded). Another thing which I consider inferior to patches by mail are pull requests. Honestly, it's very rare that I can accept a patch without some reformatting or minor corrections. I also never ask people to do such minor reformatting just to be able to click 'accept'. Again, pull requests cannot be worked-on or accepted offline. Similarly, my patches sent to other projects are similarly handled by other authors.

beagle3 13y ago
Fwiw, pull requests are merely branches im the repository (named pr-* if i'm not mistaken)

rsgalloway 13y ago

heh, what at the odds: two libs that give you object-oriented access to git repos both coincidentally named the exact same thing: python: github.com/rsgalloway/grit ruby: github.com/mojombo/grit

szhu 13y ago

It looks there are two main reasons to have a private Git server: to archive your own code, and to have control over how visitors and collaborators view your shared code. If you want a place to store or archive personal projects (because you don't want to pay for private Git hosting?) features like wikis and issues become rather irrelevant, and if you want to move that data to or from GitHub, that takes some effort to be explicitly exported instead of just adding a new remote. If you're trying to keep your own code, running your own Git server is simple if you have SSH (or not even SSH if you want to create a local "server"). The Git website has a simple tutorial for this: http://git-scm.com/book/en/Git-on-the-Server-Setting-Up-the-Server http://git-scm.com/book/en/Git-on-the-Server-Setting-Up-the-... I've also simplified this workflow into a tiny command, "mygit": https://github.com/interestinglythere/mygit https://github.com/interestinglythere/mygit As for running a GitHub clone, unless you have some really edge-case needs or you really don't want to pay for private collaborative repos, your time is probably better spent maintaining your actual code. In addition, having your code on a social site will make it more receptive to contributors. If you don't want contributors, why do you need a /Git/ server? (If it's because that's how you like to manage your code, see my first point.)