consul does not resolve CNAME to A-record for local service

[nilsotto]

When registering an external service in the catalog (https://www.consul.io/docs/guides/external.html), consul returns a CNAME-record on the dns-interface.
Adding the recursors config option makes consul resolve the A-record an return that as well.

However, consul should always resolve CNAMEs for internal services.

example
consul 0.5.2, started like this :
consul agent --client 0.0.0.0 --server --data-dir . --bootstrap-expect 1

(Using httpie as a curl alternative:)

$  http PUT :8500/v1/agent/service/register << EOF
 {
   "ID": "srvtest1",
   "Name": "srvtest1",
   "Tags": [
     "master",
     "v1"
   ],
   "Address": "127.0.0.1",
   "Port": 8000
 } 
EOF

$ http PUT :8500/v1/catalog/register << EOF
 {
    "Datacenter": "dc1", 
    "Node": "aliastest",
    "Address": "srvtest1.service.consul",
    "Service": {
            "Service": "srvalias1", 
            "Port": 80
            }
 }
EOF

This registers the internal service srvtest1.service.consul and the alias srvalias1.service.consul

When querying the consul DNS for the alias, i would expect the CNAME to be resolved, but it is not.

$ dig @localhost -p 8600 srvalias1.service.consul 
---- snip ----
;; ANSWER SECTION:
srvalias1.service.consul. 0 IN  CNAME   srvtest1.service.consul.

However, when I start consul with the parameter --recursor localhost:8600, the CNAME is resolved to its correct A record

$ dig @localhost -p 8600 srvalias1.service.consul 

---- snip ----
;; ANSWER SECTION:
srvalias1.service.consul. 0 IN  CNAME   srvtest1.service.consul.
srvtest1.service.consul. 0  IN  A   127.0.0.1

expected behaviour
Consul should consider if the external service is on the local domain and resolve CNAME->A accordingly without consiguring a recursor,

[DanyC97]

+1

[cdosso]

Enabling --recursor localhost:8600 gives me tons of errors like this in logs:

2015/10/05 15:16:12 [ERR] dns: recurse failed: read udp 127.0.0.1:8600: i/o timeout

Btw cname resolution works but with 100% CPU usage 👎 Any idea?

[slackpad]

Hi @cdosso can you provide an example that gets into this state? Your issue should probably be filed separately from this one, which looks like an enhancement.

[goacid]

+1 @nilsotto
Adding --recursor do not fix for me :
Consul v0.5.2
Consul Protocol: 2 (Understands back to: 1)

[Nomon]

+1

[goacid]

label enhancement seems not appropriate. Without recurse DNS is not usable in many cases

[slackpad]

@goacid that's fair - changing to a bug.

[nitram509]

+1

[ghost]

+1

[waswrongassembled]

+1

[dnltsk]

+1

[Solisol]

Any progress on this issue?

[kesselborn]

same question here: any progress on this? Would be nice to remove the dnsmasq hack we currently have to use :)

[n2taylor]

Any good workarounds for this issue that do not involve enabling recursor mode?

[kyhavlov]

Fixed in #2444

[vtahlani]

+1