FQDN of host is a reverse lookup, not the hosts configured FQDN

[rolffujino]

socket.getfqdn doesn't return the configured FQDN of the host but the DNS reverse lookup of the host IP.

This can cause problems and headaches if the DNS is wrong, it would be better to rely on what is actually configured on the host, since it is supposed to be reporting on the host.

OS: Ubuntu 22.04
Python: 3.10.12

[lbr38]

Hello

I am not sure to understand. How could I reproduce this? Can you please give results of:

hostname

hostname -f

hostnamectl status

cat /etc/hosts

[rolffujino]

I will send the requested details when I am back at the office, however the key to replicating the issue is to have a PTR record on the DNS server that returns the wrong FQDN.

For example, the IP address and hostname of the node is:

• 10.0.0.1
• test.prod.local

The DNS server has an incorrect PTR record that resolves 10.0.0.1 to foobar.prod.local.

When you use the socket.getfqdn() function it will return the foobar.prod.local name and not the test.prod.local address.

The agent should prioritize the hosts info over data stored in DNS.
This can be achieved using the function socket.gethostname().

This is practical for the following reasons:

• When you are rapidly redploying a large set of hosts, if the reverse lookup updates or replication on the DNS server, you have a large number misaligned entries in repomanager
• The person registering the hosts may not also have access to the DNS to clean any bad records, so it could a while before the correct is displayed.
• When troubleshooting issues with the host, repomanager and linupdate, it is easier when you limit the scope to as few systems as possible.

[lbr38]

Okay much more clear thanks.

However, there is often confusion between hostname and FQDN. For me, your node hostname is test and its FQDN is test.prod.local

The reason I prefer that linupdate uses socket.getfqdn() is that each machine is properly identified with its full name in Repomanager and also to avoid errors when sending emails.

The problem with socket.gethostname() is that it does not return the FQDN, but only the machine name test.

If you have several nodes with the same hostname but different environment registered in Repomanager, for example:

• app.prod.local
• app.preprod.local
• app.sandbox.local

Changing socket.getfqdn() to socket.gethostname() will cause them to be registered with the same name in Repomanager:

• app
• app
• app

This issue may not happen if you configured your hostname with the FQDN, but it is not the case for all users...

I will need more time to think about it and find a solution that works for everyone.

[lbr38]

I changed socket.gethostname() to socket.getfqdn() one year ago because I was facing this problem with nodes having the same name in Repomanager

#63