Coredump

I’ve worked with lots of servers, most of which I cannot access directly, so I often use remote access: Remote Desktop Connection or Terminal Services in Windows, and ssh in *nix.

While I do have PasswordSafe to remember all those passwords, I’m the lazy admin type, so I often opt for password-less authentication using public keys. For this, I use PuTTY, et. al.

1. First, I generate an RSA key using PuTTYgen. I don’t enter anything for the passphrase. Warning: NOT recommended for production servers!
2. I save the public and secret keys (in .PPK format) in a directory. I also cut and paste the RSA string in a text file, key.txt.
3. For now, the remote box is configured for “normal” ssh, that is, through password authentication. So, I copy over the RSA string file to the remote box:

   C:\> pscp \path\to\key.txt user@remote-host:/home/user/.

I log on to the remote host using PuTTY, and do the following:

$ mkdir .ssh
$ mv key.txt .ssh/authorized_keys
$ chmod 700 .ssh; chmod 600 .ssh/authorized_keys

4. I also edit the SSHd config file, /etc/ssh/sshd_config, with the following parameters:

   RSAAuthentication yes
   PubkeyAuthentication yes
   AuthorizedKeysFile     .ssh/authorized_keys
   PasswordAuthentication no
   PermitRootLogin no

5. SSHd is then restarted.
6. Back in Windows, I launch PuTTY, and enter the host name for the remote box. In the SSH/Auth category, I place the private key file saved earlier. I save the session, named remote-nopass.
7. I then launch Pageant, drill down to Saved Sessions > remote-nopass. It will bring up the SSH login page, where I enter my login name, after which, I am authorized through the pubkey, and log in to the SSH session.

Seems tedious at first, but I can then export key.txt to other remote servers, and just save sessions for Pageant’s use. Pageant is conveniently located in the system tray within reach.

Here at my new workplace, there are more than 20 web apps that are being developed in-house. They go through the following workflow: devs use development and staging servers for tests, and after QA, the new builds are pushed to the production servers.

They’ve employed a custom version control system for coding, but it’s not panning out. They need an automated way of pushing updates to the dev and staging servers for a quicker turn-around time.

Enter Subversion. I’ve used Subversion both in production and in personal projects. (I haven’t used CVS so I can’t compare the two.) The way I see it, the web apps can be pushed this way:

1. Install SVN with WebDAV support.
2. Create the repositories’ directory trees:

   mkdir /path/to/repository/projectname

3. Note that the repositories have to be owned and writeable by the httpd user:

   chown -R httpd:httpd /path/to/repository/projectname; \\
   chmod g+s /path/to/repository/projectname

4. Create the SVN repositories:

   svnadmin create /path/to/repository/projectname

5. Import the current working directories for the projects:

   svn import /path/to/working/project/dir \\
   file:///path/to/repository/projectname

6. Remove the working directories (backup first):

   rm -rf /path/to/working/project/dir

7. Check out directory trees from the repo:

   svn co file:///path/to/repository/projectname

8. Edit and commit changes:

   svn ci

9. Or, pull changes:

   svn update /path/to/working/project/dir

Easy so far. I’ve also added an HTTP URL for the SVN repos using Apache’s WebDAV extensions. In httpd.conf, just add the following:

LoadModule dav_svn_module modules/mod_dav_svn.so
<Location /repos>
   DAV svn
   SVNParentPath /path/to/repo/root
</Location>

Restart the service. Now the repo can be accessed via http://reposerver/repos/projectname.

But I hit a snag: I can’t get the post-commit hook to run on commits! I’m leaning on problems with the permissions, but it’s eluding me right now. *sigh*