Up to now the home server project and articles were based on Arch Linux. Before I wrote these articles I had mostly used Slackware, but for this project I used Arch because it was more flexible at that time. But things change over time, and not always for the better. The first problems with Arch started when they abandoned the net-tools package. The scripts that should replace this functionality are still not good enough for this server project and possible for a lot of other projects.
If you use fwbuilder to configure the firewalls on your server, then don’t upgrade to iptables 1.4.16 but keep 1.4.15. Version 1.4.16 does work, but you will see a lot of warnings like:
WARNING: The state match is obsolete. Use conntrack instead.
The syntax of several iptables and ip6tables commands will change, and fwbuilder needs an upgrade so that the correct rules will be generated again. It’s just a matter of time until it really doesn’t work anymore. To stop upgrading, add IgnorePkg = iptables in /etc/pacman.conf until a fwbuilder will be released that generates the new syntax. The iptables package only depends on glibc, and as long as there is no major glibc update you can safely use the 1.4.15 version.
About a year ago I wrote an message about the deprecation of net-tools. Finally after a little more then a year the “new” scripts can do almost anything what the old scripts could do. The Arch Wiki is not yet up to date, but at least the network configuration now works for regular servers. It only doesn’t work on the Home Server itself if you use a virtual gateway server. Then you must still use the old network-legacy package that you can download from this site.
I have updated the articles that used the network-legacy method as temporary solution.
Since 15 July the old Arch Linux installer CD image is replaced. There are no more menu’s and if you don’t install Arch every day you need a guide next to you to successfully install Arch. Is this a bad thing? I don’t think so because you now have full control over the installation. From the start everything is downloaded from the Internet so you don’t need to upgrade your new installation right away. Once you have done this new installation you will see the power of simplicity.
Since CUPS v1.6.0 the network protocol changed so that printer browsing is removed. This means that if you were used to connect to network printer with a Linux client, this doesn’t work anymore. You now have to manually connect printers. The server now publishes the printers using zeroconf, so you need a working avahi server running next to the cups server.
I didn’t had much time for this site lately, a new web-site for the Dutch D-Star community was needed because one of their members passed away. The original site was generated with MS-Office Word so the source pages were almost unreadable. Most pages scored over 500 errors on the W3C Markup Validation Service.It took a week of copying and pasting to move the data to the new server. The site still needs some cleanup but for now we are happy with the result.
Another new site is about my old sound studio which I had 30 years ago with a friend.This old site needed a facelift, so this was a good moment to convert that site to WordPress and use a different theme. This site has a public and private part.
The dovecot configuration needed a small change too after the upgrade to dovecot 2.1.1.
Also one new article: Munin for system graphing and monitoring. There is some more about system monitoring in the waiting room.
There are two new articles added, one about how to setup a web server with nginx, and one that explains how to migrate from apache to nginx. I finally did take the time to study how to do things the “nginx” way after years of using apache for a web server. It was just a matter of learning how to run php and cgi scripts. The only thing left to do is the way the mercurial repository is served, but it works for this moment. Of course this server is now serving this site using nginx.
The site itself got some cosmetic changes. The static header is replaced with random photographs that I made around the place where I live. More will follow when I take more panorama type pictures that will to fit in the header without losing important parts. Amazing how few from 12000 pictures can be used for a header.
I think that Amanda is one of the better backup solutions for small and large systems. The backups itself are written using standard *nix tools and can be used to restore even without amanda itself. The backups can be written to real tapes, disks simulating a tape changer, and into the cloud using Amazon S3 storage.
Because using amanda can be complicated and simple at the same time, I have split the whole story in three articles. The first is about the server part, which is also good for single machines. The second adds a *nix client, and the last article describes some practical commands such as maintenance and restore data.
You may expect an article about Amanda Recovery Manager, this is a backup solotion for your MySQL database and is a better alternatice then what I have written in the MySQL article.