Why use IP and not domain name?

[farzadpanahi]

I was just wondering why do we need to configure our graylog server end point by IP using hostip rather than having an attribute like graylogserver and set the domain name or IP, like graylogserver="graylog.mycompany.com" ?

[oseymen-apigee]

To my mind, this is configuration because it is not something that changes during runtime and something that needs to be configured by people other than devs outside the code (e.g. infrastructure/operations guys deploying this on-premise)

Have I understood the question correctly?

[farzadpanahi]

Let me rephrase.
Basically as you know there are two ways to identify a node on a network:

• IP address (10.10.10.10)
• Domain name (graylogserver.mycompany.com)

Out of these two ways you chose the first one to identify the target graylog server. My concern is that it is usually a better practice (under most environment not all) to identify network servers by domain name rather than IP address. This is to break the dependency on hard-coded IPs and instead to use some dynamic mechanism like DNS.

So in this case for instance, the IP address for the graylog server may change but the domain name is less likely to change. That's why I was asking why not use an attribute like HOST that can accept both domain name and IP address to configure the target. This way the user can input either the IP address or the domain name of the graylog server.

I just thought that would be more convenient :>

Btw, I forked your project to add TCP support. I don't know what is it that I am missing, but it just doesn't work. I get EOFException from GZIPInputStream on the server. Have you ever tried to write the TCP support? Do you have any clue what might be going wrong?

[chrisdcarpenter]

I agree with +farzadpanahi. For me its especially important (since my graylog server is in AWS I do not have a static ip if the graylog server get redeployed). I was able to add support for DNS by modifying the Send method in the UDPTransport class to have a try/catch around the IPAddress.Parse statement with a Dns.GetHostAddress().First() in the catch. Might not be the best way to implement it, but it was fast.