In Cloudera deployments on public cloud, one of many key configuration parts is the DNS. Get it flawed and your deployment might turn out to be wholly unusable with customers unable to entry and use the Cloudera information providers. If the DNS is ready up much less best than it might be, connectivity and efficiency points might come up. On this weblog, we’ll take you thru our tried and examined greatest practices for organising your DNS to be used with Cloudera on Azure.
To get began and offer you a really feel for the dependencies for the DNS, in an Azure deployment for Cloudera, these are the Azure managed providers getting used:
- AKS cluster: information warehouse, information engineering, machine studying, and Knowledge stream
- MySQL database: information engineering
- Storage account: all providers
- Azure database for PostgreSQL DB: information lake and information hub clusters
- Key vault: all providers
Typical buyer governance restrictions and the impression
Most Azure customers use non-public networks with a firewall as egress management. Most customers have restrictions on firewalls for wildcard guidelines. Cloudera assets are created on the fly, which suggests wildcard guidelines could also be declined by the safety group.
Most Azure customers use hub-spoke community topology. DNS servers are normally deployed within the hub digital community or an on-prem information middle as a substitute of within the Cloudera VNET. Meaning if DNS shouldn’t be configured appropriately, the deployment will fail.
Most Cloudera prospects deploying on Azure permit using service endpoints; there’s a smaller set of organizations that don’t permit using service endpoints. Service endpoint is an easier implementation to permit assets on a non-public community to entry managed providers on Azure Cloud. If service endpoints will not be allowed, firewall and personal endpoints would be the different two choices. Most cloud customers don’t like opening firewall guidelines as a result of that can introduce the danger of exposing non-public information on the web. That leaves non-public endpoints the one choice, which can even introduce extra DNS configuration for the non-public endpoints.
Connectivity from non-public community to Azure managed providers
Firewall to Web
Route from firewall to Azure managed service endpoint on the web straight.
Service endpoint
Azure supplies service endpoints for assets on non-public networks to entry the managed providers on the web with out going by the firewall. That may be configured at a subnet degree. Since Cloudera assets are deployed in several subnets, this configuration should be enabled on all subnets.
The DNS information of the managed providers utilizing service endpoints can be on the web and managed by Microsoft. The IP tackle of this service can be a public IP, and routable from the subnet. Please discuss with the Microsoft documentation for element.
Not all managed providers help providers endpoint. In a Cloudera deployment situation, solely storage accounts, PostgreSQL DB, and Key Vault help service endpoints.
Luckily, most customers permit service endpoints. If a buyer doesn’t permit service endpoints, they must go along with a non-public endpoint, which has similarities to what must be configured within the following content material.
Personal Endpoint
There’s a community interface with a non-public IP tackle created with a non-public endpoint, and there’s a non-public hyperlink service related to a particular community interface, in order that different assets within the non-public community can entry this service by the non-public community IP tackle.
The important thing right here is for the non-public assets to discover a DNS resolve for that non-public IP tackle. There are two choices to retailer the DNS file:
- Azure managed public DNS zones will all the time be there, however they retailer various kinds of IP addresses for the non-public endpoint. For instance:
- Storage account non-public endpoint—the general public DNS zone shops the general public IP tackle of that service.
- AKS API server non-public endpoint—the general public DNS zone shops the non-public IP of that service.
- Azure Personal DNS zone: The DNS information can be synchronized to the Azure Default DNS of LINKED VNET.
Personal endpoint is eligible to all Azure managed providers which are utilized in Cloudera deployments.
As a consequence, for storage accounts, customers both use service endpoints or non-public endpoints. As a result of the general public DNS zone will all the time return a public IP, the non-public DNS zone turns into a compulsory configuration.
For AKS, these two DNS options are each appropriate. The challenges of personal DNS zones can be mentioned subsequent.
Challenges of personal DNS zone on Azure non-public community
Vital Assumptions
As talked about above for the everyday situation, most Azure customers are utilizing a hub-and-spoke community structure, and deploy customized non-public DNS on hub VNET.
The DNS information can be synchronized to Azure default DNS of linked VNET.
Easy Structure Use Circumstances
One VNET situation with non-public DNS zone:
When a non-public endpoint is created, Cloudera on Azure will register the non-public endpoint to the non-public DNS zone. The DNS file can be synchronized to Azure Default DNS of linked VNET.
If customers use customized non-public DNS, they’ll configure conditional ahead to Azure Default DNS for the area suffix of the FQDN.
Hub-and-spoke VNET with Azure default DNS:
With hub-spoke VNET and Azure default DNS, that’s nonetheless acceptable. The one drawback is that the assets on the un-linked VNET will be unable to entry the AKS. However since AKS is utilized by Cloudera, that doesn’t pose any main points.
The Problem Half
The most well-liked community structure amongst Azure shoppers is hub-spoke community with customized non-public DNS servers deployed both on hub-VNET or on-premises community.
Since DNS information will not be synchronized to the Azure Default DNS of the hub VNET, the customized non-public DNS server can’t discover the DNS file for the non-public endpoint. And since the Cloudera VNET is utilizing the customized non-public DNS server on hub VNET, the Cloudera assets on Cloudera VNET will go to a customized non-public DNS server for DNS decision of the FQDN of the non-public endpoint. The provisioning will fail.
With the DNS server deployed within the on-prem community, there isn’t Azure default DNS related to the on-prem community, so the DNS server couldn’t discover the DNS file of the FQDN of the non-public endpoint.
Configuration greatest practices
Towards the background
Possibility 1: Disable Personal DNS Zone
Use Azure managed public DNS zone as a substitute of a non-public DNS zone.
- For information warehouse: create information warehouses by the Cloudera command line interface with the parameter “privateDNSZoneAKS”: set to”None.”
- For Liftie-based information providers: the entitlement “LIFTIE_AKS_DISABLE_PRIVATE_DNS_ZONE” should be set. Clients can request this entitlement to be set both by a JIRA ticket or have their Cloudera resolution engineer to make the request on their behalf.
The only real downside of this selection is that it doesn’t apply to information engineering, since that information service will create and use a MySQL non-public DNS zone on the fly. There may be at current no choice to disable non-public DNS zones for information engineering.
Possibility 2: Pre-create Personal DNS Zones
Pre-create non-public DNS zones and hyperlink each Cloudera and hub VNETs to them.
The benefit of this strategy is that each information warehouse and Liftie-based information providers help pre-created non-public DNS zones. There are nevertheless additionally just a few drawbacks:
- For Liftie, the non-public DNS zone must be configured when registering the atmosphere. As soon as previous the atmosphere registration stage, it can’t be configured.
- DE will want a non-public DNS zone for MySQL and it doesn’t help pre-configured non-public DNS zones.
- On-premises networks can’t be linked to a non-public DNS zone. If the DNS server is on an on-prem community, there are not any workable options.
Possibility 3: Create DNS Server as a Forwarder.
Create a few DNS servers (for HA consideration) with load balancer in Cloudera VNET, and configure conditional ahead to Azure Default DNS of the Cloudera VNET. Configure conditional ahead from the corporate customized non-public DNS server to the DNS server within the Cloudera subnet.
The downside of this selection is that extra DNS servers are required, which ends up in extra administration overhead for the DNS group.
Possibility 4: Azure-Managed DNS Resolve
Create a devoted /28 subnet in Cloudera VNET for Azure non-public DNS resolver inbound endpoint. Configure conditional ahead from customized non-public DNS to the Azure non-public DNS resolver inbound endpoint.
Abstract
Bringing all issues collectively, think about these greatest practices for organising your DNS with Cloudera on Azure:
- For the storage account, key vault, postgres DB
- Use service endpoints as the primary selection.
- If service endpoint shouldn’t be allowed, pre-create non-public DNS zones and hyperlink to the VNET the place the DNS server is deployed. Configure conditional forwards from customized non-public DNS to Azure default DNS.
- If the customized non-public DNS is deployed within the on-premises community, use Azure DNS resolver or one other DNS server as DNS forwarder on the Cloudera VNET. Conditional ahead the DNS lookup from the non-public DNS to the resolver endpoint.
- For the info warehouse, DataFlow, or machine studying information providers
- Disable the non-public DNS zone and use the general public DNS zone as a substitute.
- For the info engineering information service
- Configure the Azure DNS resolver or one other DNS server as a DNS forwarder on the Cloudera VNET. Conditional ahead the DNS lookup from the non-public DNS to the resolver endpoint. Please discuss with Microsoft documentation for the main points of organising an Azure DNS Personal Resolver.
For extra background studying on community and DNS specifics for Azure, take a look at our documentation for the assorted information providers: DataFlow, Knowledge Engineering, Knowledge Warehouse, and Machine Studying. We’re additionally completely satisfied to debate your particular wants; in that case please attain out to your Cloudera account supervisor or get in contact.
