dial tcp lookup image-registry.openshift-image-registry.svc on 10.10.8.132-53 no such host

[bobby0724]

I have successfully completed installation on bare metal using UPI and I have setup NFS share for the internal registry

I try to deploy a sample application and I am getting the following error

seems that the OKD is querying my DNS instead of the internal DNS service is there any way i an fix this ?

Failed to pull image "image-registry.openshift-image-registry.svc:5000/basictest/httpd-sample:latest": rpc error: code = Unknown desc = error pinging docker registry image-registry.openshift-image-registry.svc:5000: Get "https://image-registry.openshift-image-registry.svc:5000/v2/": dial tcp: lookup image-registry.openshift-image-registry.svc on 10.10.8.132:53: no such host

[bobby0724]

@fortinj66 is there a chance you can help me to check this issue

[gudroot]

To me this looks like another case of the search=dot problem from issue #648 (comment)

[fortinj66]

Is your internal registry running? I was able to recreate this on a new UPI/DHCP cluster but the issue really was that the internal registry was not configured/Running. Once that was completed internal DNS was fine...

check to see if you have a service entry for the registry

# oc get service -n openshift-image-registry
NAME                      TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)     AGE
image-registry            ClusterIP   172.30.181.130   <none>        5000/TCP    25m
image-registry-operator   ClusterIP   None             <none>        60000/TCP   103m

[bobby0724]

Is your internal registry running? I was able to recreate this on a new UPI/DHCP cluster but the issue really was that the internal registry was not configured/Running. Once that was completed internal DNS was fine...

check to see if you have a service entry for the registry

# oc get service -n openshift-image-registry
NAME                      TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)     AGE
image-registry            ClusterIP   172.30.181.130   <none>        5000/TCP    25m
image-registry-operator   ClusterIP   None             <none>        60000/TCP   103m

Yes it is running actually sending the hostname and dns from DHCP fix the issue, but as you said FCOS should be able to resolve the name

at some point i end up with all the nodes with localhost as hotname and nothing works, after setting hostname in the DHCP reservations everything go back to normal

[fortinj66]

Yep, don't depend on reverse DNS host assignment...

[fortinj66]

OK, I can reproduce with static IP addresses configured on a vanilla FCOS 34 install

[core@worker0 ~]$ cat /etc/resolv.conf | grep -v "\#"

nameserver 10.99.111.1
nameserver 10.99.111.2
search .

[fortinj66]

Potential workaround/fix for 'search .' issue

openshift/okd-machine-os#156