Rudder

It could be an other instance of http://www.rudder-project.org/redmine/issues/3924

Does comments on that ticket seems to feet your use case ?

François ARMAND wrote:

It could be an other instance of http://www.rudder-project.org/redmine/issues/3924

Does comments on that ticket seems to feet your use case ?

Hi François,

What do you check :

1/ The IP address is unique ?

2/ The IP address is in the allowed network range ?

3/ ?

In our case all the the hosts have and unique UUID and the same IP address.

Thanks

The UUID is not taken into account here.

So, to answer your question, we are checking the uniqueness of the IP address. If you have two nodes with exactly the same IP, CFEngine server won't be able to know what is the hostname of the node. Or all your node also have the same hostname ? If that is the case, I'm pretty sure we don't support that use case for now, I will need the help of a more ops guy to help understand it.

Well, there is one case that I think we missed: in the case of relay server (each relay server can talk to a local net), it is completly legit to have two nodes with the same ip and different hostname. So we should check the uniquness of the couple (ip, policy server id).

The correction here is to remove the special check that we added in Rudder UI to forbid duplicated IPs in authrozied networks.

This not sufficient to allow a working NATted network, see: #4442

i guess it's too big a change of behaviour to be changed in a stable version (we are basically moving from : you can't have twice the same ip, which is convenient if you duplicated VMs and don't check their IP configuration) to you can have twice the same IP

Not that relay server where introduced in 2.8, so there can't be cases in 2.6 when duplicate ip would make sense

Applied in changeset 0141500bf2ca9b26e74e9a2206d3c0ac6d6261df.