Ubuntu – Change Exim4 www-data user (Ubuntu)

Attempting to expand on @Zoredache's answer, as I give this a go myself:

• Create a new group (www-pub) and add the users to that group

  groupadd www-pub

  usermod -a -G www-pub usera ## must use -a to append to existing groups

  usermod -a -G www-pub userb

  groups usera ## display groups for user

• Change the ownership of everything under /var/www to root:www-pub

  chown -R root:www-pub /var/www ## -R for recursive

• Change the permissions of all the folders to 2775

  chmod 2775 /var/www ## 2=set group id, 7=rwx for owner (root), 7=rwx for group (www-pub), 5=rx for world (including apache www-data user)

  Set group ID (SETGID) bit (2) causes the group (www-pub) to be copied to all new files/folders created in that folder. Other options are SETUID (4) to copy the user id, and STICKY (1) which I think lets only the owner delete files.

  There's a -R recursive option, but that won't discriminate between files and folders, so you have to use find, like so:

  find /var/www -type d -exec chmod 2775 {} +

• Change all the files to 0664

  find /var/www -type f -exec chmod 0664 {} +

• Change the umask for your users to 0002

  The umask controls the default file creation permissions, 0002 means files will have 664 and directories 775. Setting this (by editing the umask line at the bottom of /etc/profile in my case) means files created by one user will be writable by other users in the www-group without needing to chmod them.

Test all this by creating a file and directory and verifying the owner, group and permissions with ls -l.

Note: You'll need to logout/in for changes to your groups to take effect!

The server appears to be doing throttling which might affect you if you are testing from one IP:

2009-08-10 07:21:05 H=(aurl.domain.ni) [127.0.0.1] Warning: Sender rate 4.6 / 1h

Often issues like this are related to poor DNS. Exim does a lot of resolution during processing mail. Check the DNS config on the Exim server. Check the performance on the DNS server it is using.

Is the server doing any DNSBL processing? I'm often asked to look at servers exhibiting poor performance and it turns out they are using DNSBLs that have been retired, meaning lookups are timing out.

Try running exim in debug on the server to see where the problem is. The following will run an Exim daemon on port 26, staying attached to the console and printing lots of debug:

exim -bd -d -oX 26

Then telnet to tcp/26 instead of tcp/25 and test as you had been.