Bind Operations - Set Up DNS Server

From NovaOrdis Knowledge Base
Jump to navigation Jump to search

External

Internal

Overview

This is a step by step guide on installing and configuring a bind DNS server. We needed it during the deployment of an OpenShift environment. During this procedure, we will create a new "openshift35.local" zone file and all OpenShift hosts will get names within the "*.openshift35.local" DNS domain. Additional zone files/local domains can be added later.

Relevance

  • RHEL 7.1
  • Centos 7.3

Binary Installation

Intall bind binaries:

yum install bind* -y

DO NOT yum remove bind*, it will remove dhclient and related libraries, and that might render your host inoperable.

Starting and Stopping the Server

Installing binaries will deploy a bind server with a basic configuration that can be started and stopped.

Starting the Server

You can start the server right away, it should start fine. See Troubleshooting the Server.

systemctl start named

Stopping the Server

systemctl stop named

Startup Options

named startup options are specified in /etc/sysconfig/named as follows:

OPTIONS="-4 -d 10"

where:

-4

Instructs the server to use IPv4 only even if the host machine is capable of IPv6. -4 and -6 are mutually exclusive.

-6

Instructs the server to use IPv6 only even if the host machine is capable of IPv4. -4 and -6 are mutually exclusive.

-d <debug-level>

Specifies the debug level. Usually 10l.

Edit /etc/named.conf

/etc/named.conf is the main configuration file.

Use the default and modify the following:

listen-on

Specifies the IPv4 network interface on which to listen for queries.

On a DNS server that acts as a gateway, you can use this option to answer queries originating from a single network only. All IPv4 interfaces are used by default.

options {

        ...

        listen-on port 53 { any; };

        ...

}

Alternatively, to specify a single interface:

        listen-on port 53 { 172.30.0.3; };

or several:

        listen-on port 53 { 127.0.0.1;172.30.0.3; };

listen-on-v6

Specifies the IPv6 network interface on which to listen for queries.

On a DNS server that also acts as a gateway, you can use this option to answer queries originating from a single network only. All IPv6 interfaces are used by default.

options {

        ...

        listen-on-v6 port 53 { any; };

        ...

}

To prevent listening on IPv6 ports:

        listen-on-v6 port 53 { none; };
}

allow-query

Specifies which hosts are allowed to query the nameserver for authoritative resource records.

It accepts an access control list, a collection of IP addresses, or networks in the CIDR notation. All hosts are allowed by default.

options {

        ...

        allow-query     { any; };

        ...

}

Alternatively, an "acl" can be specified as follows:

...

acl allowed-to-query {
        172.30.0.0/16;
        localhost;
        localnets;
};

options {

       ...
       allow-query     { allowed-to-query; };
      
       ...
}

Adding a Zone

Add a zone in /etc/named.conf, as follows. Conventionally, place it between the zone "." declaration and "include", as shown below.

Note that more than one zone file can be declared here.

In case of sub-domains, a separate zone file should be added for each sub-domain, as shown below for "openshift.external" and "apps.openshift.external".

...
zone "." IN {
        type hint;
        file "named.ca";
};

zone "openshift.local" IN {
        type master;
        file "openshift.local.zone";
};

zone "openshift.external" IN {
        type master;
        file "openshift.external.zone";
};

zone "apps.openshift.external" IN {
        type master;
        file "apps.openshift.external.zone";
};

include "/etc/named.rfc1912.zones";
...

The zone file path specified after "file" is relative to /var/named.

Start with named.localhost as a model:

cd /var/named
cp named.localhost openshift.local.zone

Make sure the file has the same ownership as the rest of the files in the directory:

chown root:named openshift.local.zone

Assigning the ownership is critical. The file won't be read and names from the file will not be resolved otherwise.

Zone File

/var/named/openshift.local.zone:

$TTL 3600
@                       IN SOA  openshift.local. root (
                                        2015092101      ; serial
                                        3600            ; refresh
                                        15M             ; retry
                                        1W              ; expire
                                        1D )            ; minimum
                        IN NS   localhost
localhost               IN A    127.0.0.1
ns                      IN A    172.20.2.10
openshift-master1       IN A    172.20.2.1
openshift-node1         IN A    172.20.2.2
openshift-node2         IN A    172.20.2.3

PTR (Inverse Record) File

TODO: not tested.

/var/named/openshift.local.rzone:

$TTL 3600
@			IN SOA	openshift.local. root.openshift.local. (
					2015092101	; serial
					3600		; refresh
					15M		; retry
					1W		; expire
					1D )		; minimum
			IN NS	localhost.
172.20.2.1		IN PTR 	openshift-master1.openshift.local.
172.20.2.2		IN PTR  openshift-node1.openshift.local.

Make sure the file has the same ownership as the rest of the files in the directory:

chown root:named openshift.local.rzone

Assigning the ownership is critical. The file won't be read and names from the file will not be resolved otherwise.

I think the reverse resolution is not fully enabled until I add something like the following in /etc/named.conf:

zone"0.168.192.in-addr.arpa" IN {
type master;
file "reverse.linuxzadmin";
allow-update { none; };
};

Verify Configuration Files

named-checkconf /etc/named.conf 

named-checkzone  <zone-name> <zone-file>
named-checkzone  openshift.local.zone /var/named/openshift.local.zone 

Configure iptables

Add the following rules to your iptables configuration:

...
iptables -A INPUT -p tcp -m state --state NEW -m tcp --dport 53 -j ACCEPT
iptables -A INPUT -p udp -m state --state NEW -m udp --dport 53 -j ACCEPT
...

For more details on the iptables configuration, see

iptables IPv4 Base Configuration

Make sure the correct rules are in place with iptables -F:

Also see

ping and ssh succeed but telnet fails

Configuring Recursive Resolution

If a name cannot be resolved by the current bind server, the query is forwarded to parent DNS servers. This makes the DNS server a recursive DNS server.

Recursion is enabled by using "recursion" keyword:

options {

     ...
     recursion yes;
     ...
}

If the recursive DNS server has a public IP address, then access control MUST be enabled to limit queries to legitimate users. Failing to do so will cause the server to become part of large scale DNS amplification attacks.

Domain Name System Security Extensions (DNSSEC) Configuration

options {

     ...
     dnssec-enable yes;
     dnssec-validation yes;
     dnssec-lookaside auto;

     /* Path to ISC DLV key */
     bindkeys-file "/etc/named.iscdlv.key";

     managed-keys-directory "/var/named/dynamic";
     ...
}

Declared a Wildcard DNS Domain

A wildcard DNS domain or sub-domain is a domain that resolves all the host and subdomains under it to the same IP address.

Note that the sub-domains are logically hierarchical, a different zone file is required for each sub-domain. The zone files are separate entities and do not have a hierarchical structure. For more details see

Domains and Sub-domains


This is an example of a zone file for the "apps.openshift35.external" wildcard subdomain. The content goes to /var/named/apps.openshift35.external.zone:

$TTL 3600
@	  IN SOA	apps.openshift35.external. root (
					2017052401	; serial
					3600		; refresh
					15M		; retry
					1W		; expire
					1D )		; minimum
	  IN 	NS localhost
	  IN    A  172.23.0.12

localhost IN 	A  127.0.0.1

*	  IN	A  172.23.0.12

Make sure the file has a "root:named" ownership:

chown root:named apps.openshift35.external.zone

Assigning the ownership is critical. The file won't be read and names from the file will not be resolved otherwise.

The /etc/named.conf declaration of the zone file:

...

zone "apps.openshift35.external" IN {
        type master;
        file "apps.openshift35.external.zone";
};

...

Starting at Boot

systemctl enable named

Test:

systemctl is-enabled named
enabled

Configure The Local DNS Resolver

Same as configuring clients on other machines in the network. See:

Set Up DNS Client

Configure DNS Resolver on Other Hosts

Set Up DNS Client

Reboot and Test

Local Tests

Assuming that the name of the name server is "ns.openshift.local", and "openshift-master1.openshift.local" is a valid host in domain, after reboot, the following must be successful:

# nslookup ns
Server:		172.20.2.10
Address:	172.20.2.10#53

Name:	ns.openshift.local
Address: 172.20.2.10
# nslookup ns.openshift.local
Server:		172.20.2.10
Address:	172.20.2.10#53

Name:	ns.openshift.local
Address: 172.20.2.10
# nslookup openshift-master1
Server:		172.20.2.10
Address:	172.20.2.10#53

Name:	openshift-master1.openshift.local
Address: 172.20.2.1
# nslookup openshift-master1.openshift.local
Server:		172.20.2.10
Address:	172.20.2.10#53

Name:	openshift-master1.openshift.local
Address: 172.20.2.1

Also a random (but valid) name must be resolved

# nslookup www.cnn.com
Server:		172.20.2.10
Address:	172.20.2.10#53

Non-authoritative answer:
www.cnn.com	canonical name = turner.map.fastly.net.
Name:	turner.map.fastly.net
Address: 23.235.47.73

Remote Tests

Adding a New Host to the Zone File

Go to the corresponding zone file (example if the domain is openshift.local, the zone file is /var/named/openshift.local.zone)

...
master31-1              IN A    172.20.2.11

Also modify the inverse record file /var/named/openshift.local.rzone (if exists)

172.20.2.11		IN PTR  master31-1.openshift.local.

Then reboot the name server.

After reboot, run nslookup from the name server itself, an old host and the newly setup host:

[openshift@ns ~]$ nslookup master31-1.openshift.local
Server:		172.20.2.10
Address:	172.20.2.10#53

Name:	master31-1.openshift.local
Address: 172.20.2.11

Troubleshooting the Server

systemctl status named

Logs are generated in two places, and the content is different:

journalctl -f -u named
tail -f /var/named/data/named.run

Debug verbosity can be changed by adding the following in /etc/sysconfig/named:

OPTIONS="-d 10"

Troubleshooting Situations

Network Unreachable

error (network unreachable) resolving 'www.cnn.com/A/IN': 2001:7fe::53#53

Got this on a network that prevented UDP DNS traffic. The server resumed working find immediately after I connected to a different network.