No name resolution at advertise parameter

[andienase]

I understand the problem with the private addresses for the bind parameter so I start the consul agent with the -bind paramter but with this I get an error for the advertise address

-> consul agent -node consul-s1-bind consul-s1 -config-file /home/consul/cfg/consul.cfg
==> WARNING: It is highly recommended to set GOMAXPROCS higher than 1
==> Starting raft data migration...
==> Starting Consul agent...
==> Error starting agent: Failed to start Consul server: Failed to start lan serf: Failed to parse advertise address!

The documentation says:
-advertise - The advertise address is used to change the address that we advertise to other nodes in the cluster. By default, the -bind address is advertised.

So ok I tried to tell him the same address as the bind address but it still not working:

-> consul agent -node consul-s1 -bind consul-s1 -advertise consul-s1 -config-file /home/consul/cfg/consul.cfg
==> WARNING: It is highly recommended to set GOMAXPROCS higher than 1
==> Starting raft data migration...
==> Starting Consul agent...
==> Error starting agent: Failed to parse advertise address: consul-s1

The name consul-s1 is resolved via LDAP and is pingable on the os (Linux/SLES 11). The node only starts when I commit the ip address to the advertise parameter.

-> consul agent -node consul-s1 -bind consul-s1 -advertise -config-file /home/consul/cfg/consul.cfg

[aidanhs]

Doesn't this require consul to be aware of NSS in order to be able to find the address on LDAP? Why not just use -advertise $(dig +short consul-s1)?

[highlyunavailable]

"Address" in this case is shorthand for "IP address" since Consul doesn't presuppose the existence of other DNS servers. @aidanhs method would work, or look up the address via some other method like checking the interfaces.

[andienase]

Sorry in the last command line the ip address after -advertise was missing.

-> consul agent -node consul-s1 -bind consul-s1 -advertise 10.10.10.10 -config-file /home/consul/cfg/consul.cfg

(ip only for example because I have currently no access to server)

What surprises me is that the name resolution is working at the -bind parameter? The 'problem' also exists when the hostname is configured in /etc/hosts

[slackpad]

I'm going to close this out - we have documented these as addresses and don't intend to support host lookups in here (even if the path in Go happens to support them).

[mitom]

Would this have an effect in case servers changed IP addresses? As in, does this effectively control how do consul nodes reach out to each other (the member list)?

[Alexey-Tsarev]

Yes, if IP changed (because of dhcpclient, a point-to-point protocol, for instance pppoe), then Consul still listening old IP address. This is true when -bind=IP provided.
When -advertise provided, Consul listened on all interfaces:

netstat -nlp | grep consul
tcp        0      0 127.0.0.1:8500          0.0.0.0:*               LISTEN      5825/consul
tcp        0      0 127.0.0.1:8600          0.0.0.0:*               LISTEN      5825/consul
tcp6       0      0 :::8300                 :::*                    LISTEN      5825/consul
tcp6       0      0 :::8301                 :::*                    LISTEN      5825/consul
tcp6       0      0 :::8302                 :::*                    LISTEN      5825/consul
udp        0      0 127.0.0.1:8600          0.0.0.0:*                           5825/consul
udp6       0      0 :::8301                 :::*                                5825/consul
udp6       0      0 :::8302                 :::*                                5825/consul

But, anyway other Consul servers can not connect to this server.

So pointed solution:
#2563
doesn't work in this case - it's necessary to restart Consul with changed IP address.

A solution is a wrapper, which periodically checks interface's IP address and restart Consul when IP changed.
I don't see any other solution. And I think, that Consul should catch IP address changes.

[slackpad]

Consul handling IP address changes is tracked under #1580.