DNS Record Management
DNS Naming Convention
A sophisticated DNS naming convention leads to intuitive and concise names and supports the use of wildcard certificates, which in turn simplifies certificate management.
The prerequisite to creating a DNS naming convention is to decide which element roles and services require a DNS name.
Forward DNS Lookup
This section summarizes DNS aspects of common element management patterns briefly, without discussing the advantages and disadvantages of each management pattern in general. A sustainable DNS provisioning solution must support all common management patterns.
Out-of-Band Management
Out-of-Band management establishes a dedicated management network to manage the network elements. Each element is connected via the management interface to this network. The management system also needs to be connected to the management network to access the network elements.
The DNS lookup resolves the IP address of the management interface.
In-Band Management
In-band management leverages the transport network, which conveys the customer traffic, and also for network management.
The DNS lookup resolves the transport layer loopback IP address for all elements connected with more than one interface to the transport network. For elements with a single connection to the transport network, the IP address assigned to this interface is resolved.
API Gateway
An API gateway forms a single entry point to access multiple elements. The management system interfaces with the API gateway and the API gateway forwards the API call to the appropriate element. The API gateway is either accessed in-band or out-of-band.
The API gateway addresses non-functional requirements such as TLS termination, authentication and authorization.
Many instances of an API gateway are typically started at the same time to avoid creating a single point of failure. All instances are accessible through the same IP address, the cluster IP address. Consequently, the API Gateway DNS name resolves the cluster IP address. Additional DNS records per gateway instance are required to access a gateway instance for troubleshooting.
The management system needs to know the REST API endpoint URL for every element. The REST API invocation is the same for all management patterns outlined before. A management process reads the REST API endpoint from the element record in the resource inventory to invoke the API. Looking at the API gateway pattern, an element does not have to have a DNS name despite providing a REST API endpoint.
Reverse DNS Lookup
A reverse DNS lookup discovers the hostname for a given IP address. Reverse DNS lookups are handy for troubleshooting as they translate an IP address to a human-friendly name. Consequently, a hostname needs to be assigned to the transport layer loopback IP address or the IP address of the interface connected to the transport network respectively.
DNS Naming Service
The DNS naming service implements the DNS naming convention and assigns a DNS name to all elements that are supposed to have a DNS name.
DNS Naming Service is beyond the scope of this document as it depends on the DNS Naming Convention. |
The resource inventory will be enhanced to store zero, one or more DNS records per element as illustrated below. The DNS Naming Service maintains the DNS records per element.
A DNS record basically consists of:
-
DNS name
-
IPv4/v6 address
-
status flag indicating whether the DNS record is enabled
-
creation date and last modification date and
-
an optional expiry date (defaults to never expire)
The resource inventory fires a domain event when a DNS record set has been added, modified or removed (DnsRecordSetModifiedEvent).
The following information is conveyed with a DNS record domain event:
-
event_id - the event ID in UUIDv4 format.
-
event_name - the event name (DnsRecordSetModifiedEvent).
-
group_id - the element group ID in UUIDv4 format
-
group_type - the element group type (set to pod)
-
group_name - the element group name
-
element_id - the ID, in UUIDv4 format, of the element whose DNS name has been modified
-
element_name - the name of the element whose DNS name has been modified
-
element_alias - an optional alias of the element whose DNS name has been modified. This property is omitted if no alias has been set.
-
element_role - the role of the element whose DNS name has been modified
-
group_id - the ID of the group the element is a member of
-
group_name - the name of the group the element is a member of
-
group_type - the type of the group the element is a member of, which is always set to pod
-
dns_recordset - the DNS record set
-
dns_zone_id - the DNS zone ID in UUIDv4 format of the DNS zone in the resource inventory.
-
dns_zone_name - the canonical DNS zone name
-
dns_name - the canonical DNS name that shall be stored in the DNS
-
dns_ttl - the optional time-to-live for the DNS record (in seconds).
-
dns_withdrawn_name - the DNS name that shall be removed from the DNS
-
dns_type - the DNS record type (e.g. A, AAAA, CNAME)
-
dns_record - the array of DNS records
-
dns_value - the DNS record value (e.g. IPv4 address, IPv6 address, alias)
-
dns_setptr - a flag indicating whether to create a PTR record
-
disabled - a flag indicating whether this record is disabled
-
-
In addition, the resource inventory fires an event if a DNS zone was added (DnsZoneCreatedEvent) or removed (DnsZoneRemovedEvent). A DNS zone event merely contains the zone ID and canonical DNS zone name:
-
event_id - the event ID in UUIDv4 format.
-
event_name - the event name (DnsZoneCreatedEvent or DnsZoneRemovedEvent).
-
dns_zone - the DNS zone that was subject to the reported change
-
dns_zone_id - the DNS zone ID in UUIDv4 format of the DNS zone in the resource inventory.
-
dns_zone_name - the canonical DNS zone name
-
DNS Connector
The DNS Connector subscribes to the DNS domain events outlined before using providing a REST API endpoint that accepts HTTP POST requests with the domain event as a requesting entity. The DNS connector creates and removes DNS zones according to the DNS zone events and translates the DNS record set domain event to resource record sets using the following semantics:
-
Store a resource record set in the DNS if the dns_name property is present. The dns_name property value becomes the resource record set name. The resource record set contains up to two resource records: a type A record for the IPv4 address and another type AAAA record for the IPv6 address. The connector derives the time-to-live (TTL) of the resource record set from the expiry date. The TTL is omitted if no expiry date has been specified.
-
Remove a resource record set from the DNS if the dns_withdrawn_name property is present. The dns_withdrawn_name property contains the name of the resource record set to be removed.
The DNS Connector uses the REST API provided by PowerDNS, an open-source DNS server, to maintain the DNS records. PowerDNS can either be configured as a DNS server or merely acts as a gateway and forwards all DNS changes to the actual DNS service.
Figure 6 illustrates the DNS records provisioning flow.
The DNS records are stored in the resource inventory and maintained by the DNS Naming Service application or manually through the user interface. All DNS record changes fire an event, which is stored in the Domain Event Queue. The domain even queue is a persistent and transactional queue, which means that domain events are stored for successfully committed transactions only. A webhook, which is a configurable service, forwards all DNS events to the DNS connector, which in turn maintains the resource records sets (RRs) in PowerDNS.
Sample Fabric DNS Naming Convention
This section introduces a sample fabric DNS naming convention. The DNS name tree is depicted below.
The <network-domain> is the base domain name for the entire network (e.g. lab.rtbrick.net for the RtBrick lab environment). The next level of the DNS tree differentiates between central management services (mgmt) and pods (pod). Every management service has a designated DNS name. Every element in a pod has a designated DNS name too. Moreover, certain interfaces (for example, the management interface) or services deployed on an element might also get a DNS name assigned.
The network diagram below illustrates a lab topology that consists of two pod fabrics, Bangalore and Nuremberg, each formed by four switches and managed by a fabric daemon as well as a central management system. The central management system consists of three services: the control center to manage the network, the log management system to query log messages and process alerts, and the telemetry management system to process and visualize metrics. Each switch has a unique router ID, which is equal to the transport layer loopback IP address, a management interface and runs the control daemon (ctrld) to manage the switch.
The table below lists samples DNS names of elements and services in the sample network topology:
DNS Name | Description |
---|---|
lab.rtbrick.net |
Network domain forming the top-level domain for all elements in the network |
mgmt.lab.rtbrick.net |
Subdomain for all management services in the network. |
leitstand.lab.rtbrick.net |
Control center service name. |
log.lab.rtbrick.net |
Log management service name. |
telemetry.lab.rtbrick.net |
Telemetry management service name. |
pod.lab.rtbrick.net |
Subdomain for all pods and their respective elements. |
blr.pod.lab.rtbrick.net |
Subdomain for all elements and services in pod Bangalore (BLR) |
nbg.pod.lab.rtbrick.net |
Subdomain for all elements and services in pod Nürnberg (NBG). |
fabric.blr.pod.lab.rtbrick.net |
Name of fabric in pod Bangalore (BLR) |
l1.blr.pod.lab.rtbrick.net |
Name of leaf 1 in pod Bangalore (BLR). This name resolves to the transport layer loopback IP address. |
me0.1.blr.pod.lab.rtbrick.net |
Name of the management interface of leaf 1 in pod Bangalore (BLR). This name resolves to the IP address assigned to the management interface. |
ctrld.l1.blr.pod.lab.rtbrick.net |
An alias to access the control daemon of leaf 1 in pod Bangalore. In the case of in-band management, the name is an alias of leaf1.blr.pod.lab.rtbrick.net whereas for out-of-band management the name is an alias for the management me0.leaf1.blr.pod.lab.rtbrick.net |
All remaining leaf and spine switches have similar names. Below is another example for spine 2 located in pod Nürrnberg (nbg) |
|
s2.nbg.pod.lab.rtbrick.net |
Name of spine 2 in pod Nürnberg (NBG). This name resolves to the transport layer loopback IP address. |
me0.s2.nbg.pod.lab.rtbrick.net |
Name of the management interface of spine 2 in pod Nürnberg (NBG). This name resolves to the IP address assigned to the management interface. |
ctrld.s2.nbg.pod.lab.rtbrick.net |
An alias to access the control daemon of spine 2 in pod Bangalore. In the case of in-band management, the name is an alias of spine2.nbg.pod.lab.rtbrick.net whereas for out-of-band management the name is an alias for the management me0.spine2.nbg.pod.lab.rtbrick.net |
The Domain Naming Service subscribes the following domain events in order to maintain the DNS names according to this naming scheme:
Domain Event | Action |
---|---|
ElementAddedEvent |
Create DNS records for the added element. |
ElementRenamedEvent |
Update the DNS records for the renamed element. |
ElementMovedEvent |
Update the DNS records for the moved element, i.e. the element is now in a different pod. |
ElementRetiredEvent / ElementRemovedEvent |
Remove the DNS records of a retired or removed element. An element must be in the retired state before it can be removed from the inventory. A retired element is inactive and kept in the repository for documentation purposes only. It depends on whether the DNS record shall be part of the documentation, and whether the records are removed when an element is retired or gets removed from the inventory. |
ElementIflAddedEvent |
Update the DNS record of an element if the management interface or transport layer loopback interface was added |
ElementIflModifiedEvent |
Update the DNS record of an element if the management interface or transport layer loopback interface IP address has changed |
ElementIflRemovedEvent |
Update the DNS record of an element if the management interface or transport layer loopback interface was removed |
PodRenamedEvent |
Update the DNS records for all elements in the pod. |