I left a few things unsaid in my previous post basic hostname set up, and after talking with some people about /etc/hosts, I thought a quick follow up is in order.

Firstly, my preference is to use the ethernet’s ip address to tie the host name to the FQDN, but others like to use 127.0.1.1. There’s a case where you really should use 127.0.1.1 and that’s in laptops where the ethernet interface might disappear when it’s disconnected, probably desktops with dynamic ip addresses, too. But for servers with a static ip address, take your pick, I chose to use eth0.

An example /etc/hosts for a laptop
127.0.0.1 localhost
127.0.1.1 mymailname.com myhostname


Where myhostname is the host name configured in the kernel.

An example /etc/hosts for a server
127.0.0.1 localhost
192.168.1.1 mymailname.com myhostname.mymailname.com someothername myhostname


When the kernel first boots, the host name is configured. How depends on the distribution. Ubuntu seems to use /etc/hostname. In my example /etc/hostname contains myhostname .

When you run the command hostname it does a gethostname() to get the configured host name. It doesn’t reread /etc/hostname every time, it’s stored in the kernel, presumably.

When you want the FQDN, you run hostname -f, it does a gethostname() to get the configured host name, then does a getaddrinfo("myhostname",...) to get the FQDN. Using the last /etc/hosts example, the FQDN is mymailname.com because it’s the first string after the ip address in the line where myhostname appears.

Short names such as the host name myhostname should go at the end of the line, what you want the FQDN to be should be the first string after the ip address, and any other names goes in the middle.

In my examples, I made the FQDN the mail name, because some distributions use the FQDN by default as the mail name. Ubuntu doesn’t, it appears to use /etc/mailname.

Anyway, that’s my understanding based on messing around and reading a man page. Hopefully I’ve explained how the host name and the FQDN is tied together in /etc/hosts.

I was helping out someone with a postfix issue, and something just clicked that will speed up diagnosing similar problems. It’s not so much a postfix problem as a hostname problem, and it’s very basic.

The symptom was, mail sent to the server at domain.com was being bounced. Looking at postfix logs, postfix was trying to deliver the mail to s123.s123. (something like that, anyway)

To the quick, hostname printed s123 and hostname -f said something to the tune of it couldn’t find the FQDN.

The fix, edit /etc/hosts to look like


127.0.0.1 localhost
1.2.3.4 domain.com s123


Where 1.2.3.4 is the ethernet’s ip address.

Now, hostname said s123 and hostname -f said domain.com . A restart of postfix, and it was accepting mail again.

The thing that clicked was the relationship between /etc/hosts, hostname and hostname -f.

See basic hostname set up part 2

I had a job come in last week for a linux file server to connect a few windows machines together, and a few printers. The client was exposed to a linux file server in a previous job, and wanted something similar.

It was a small business, with not much money to spend, and the server was to be located in the office, no server room. This is the type of job I love doing.

I sourced an ex corporate desktop for about $150 including shipping, I think they can be had for a little less, but this was convenient. It was a P4 with SATA, which is what I was looking for. The ugliest case you could image, but had space for two hard drives. DVD writeable drive, 1 gig memory, built in gigabyte ethernet. It was perfect.

I bought two new 500 gig green SATA drives for RAID 1 configuration, $70 each, and an external 500 gig drive for backup, which ran $100. So about $400 in hardware, but a really nice set up, should be very reliable, and the data should be very safe.

I was a bit of a cowboy when it came to the operating system, with karmic about to be released, I chose it over jaunty or LTS because it had a newer Samba. Turns out the Samba version wasn’t very important as most of the PC’s were running XP home edition, which doesn’t even do domain logins.

I tired to convince the client to get two external drives to keep one off site in case of a fire or something. But instead we are using the DVD drive to write backups for off site storage.

It was a really nice job, gave me a lot of energy and the client is happy. Can’t ask for more than that.

One of the minor issues I had since installing ubuntu on my eee pc was I couldn’t work out how to turn the wifi off. There’s a Fn+F2 key combo that works in xandros, but it didn’t work for me in ubuntu. I tried googling for the answer and got nowhere.

So, for google, to turn wifi off, I had to echo 0 > /proc/acpi/asus/wlan Turning it on is obviously echo 1 > /proc/acpi/asus/wlan

Another problem fixed (sort of), I wouldn’t mind fixing Fn+F2, but I’m happy enough with the echo for now.

We were given a server to play with to do whatever we wanted, hosted in a US data centre. It was running RHEL 4.x, I could have worked with that, but it didn’t look like it was on a support contract, I wasn’t able to update it, anyway.

I thought about trying to convert it to centos, but realised it would be more fun to upgrade it to ubuntu gutsy.

Quick steps from memory

1. swapoff -a
2. mkfs.ext3 /dev/VolGroup00/LogVol01 (old swap)
3. mount /dev/VolGroup00/LogVol01 /mnt
4. wget debootstrap.deb (from gutsy)
5. ar x debootstrap.deb
6. tar -C / -xzf data.tar.gz
7. debootstrap –arch=i386 gutsy /mnt
8. chroot /mnt
9. mount proc, edit /etc/fstab and /etc/network/interfaces
10. apt-get install ubuntu-minimal ubuntu-standard linux-image postfix openssh-server plus a few other packages, some that were recommends.
11. copied gutsy /boot/* to the real /boot and created a new entry in menu.lst
12. created an account and uploaded some ssh keys
13. probably some steps I’ve forgotten
14. reboot

Easy peasy, just waiting for it to come up, it’s been 5 hours now, still waiting. Must be the slowest booting server ever.

Guess I’ll be calling the states now. :’(