April release of Arc data services

[Guest Dinakar-Nethi]

hi folks

 

 

 

As many of you may know, we have a monthly release coinciding with the patch Tuesday. So, our April release went out on the 12th.

 

The release includes updates for both Azure extension for SQL Server as well as Arc data services. Here's a quick rundown of some features that shipped in each of these services.

 

Azure extension for SQL Server:

 

For Arc enabled SQL Server, we shipped one of the building blocks under Backups and Restore category. The extension can now perform automatic backups for all your databases on the Arc enabled SQL server. The backups are disabled by default so as not to break any existing backup routines. You can turn on automatic backups as follows:

 

 

 

--Install the arcdata extension if not already done

az extension add --name arcdata

 

az sql server-arc backups-policy set --name <arc-server-name> --resource-group <resourcegroup> --retention-days <retentiondays>

 

 

 

Now, before you configure the backups, there are certain roles that need to be added and permissions that need to be granted to this role. Full details described at Configure automated backup - SQL Server | Microsoft Learn

 

We are working on improving the permission set for future release to use a managed identity that will align with the principles of least privilege access mode.

 

The current automated backups only supports a default schedule, and at instance level, as follows:

 

• Weekly full backup
  
• Daily differential backup
  
• Transaction log every 5 minutes
  

 

The only parameter that can be configured with this release is the --retention-days. A value of 0 means no backups will be performed. Retention can be configured for up to 35 days. If you need the backups to be available for longer than 35 days, you can copy these files off the backup folder into a different folder. Again, our goal is to provide more flexibility to customers both in terms of scheduling as well as granularity, in future releases.

 

 

 

Arc SQL MI:

 

For Arc data services, in addition to some bug fixes and optimizations, we added the following features:

 

• Direct mode support for Failover Groups - Until this release, you could only create Failover Groups between two Arc enabled SQL managed instances using the "indirect" mode commands, meaning using the --k8s-namespace and --use-k8s. With this release we added support for creating the Failover Groups in direct mode as well closing the gap there. The Disaster recovery - Azure Arc-enabled SQL Managed Instance - Azure Arc | Microsoft Learn has been updated to include commands for both modes, as well as clarify the failover scenarios a bit more.
  
• Better scheduling of ha-orchestrator pods - When you deploy Arc data SQL MI, by default, only one ha-orchestrator pods is deployed, unless you specify additional replicas using the --orchestrator-replicas parameter. In the default configuration, the single ha-orchestrator pod was getting scheduled on the node hosting the primary replica. The issue with this deployment is that, if the node hosting the primary pod goes down, the orchestrator pod also goes down. So even though you have additional replicas available and fully synchronized, the applications cannot connect to it since the orchestrator that is responsible for this redirection is down. With this release, we updated the deployment routine to schedule the orchestrator pod on one of the secondary nodes. This will greatly improve the availability of your Arc SQL MI.
  
• Update to deploying Azure Arc data controller - Previously there were 2 ways to deploy the Arc data controller - (1) Individual mode - where you deploy the the k8s-extension first which will deploy the bootstrapper, then create the Custom Location and finally create the Data controller, and (2) All-in one experience where the az arcdata dc create command would deploy all the components in a single, seamless experience. One of the parameters that was passed as part of this deployment is the --auto-upgrade for the k8s-extension which we set to false as Arc data services doesn't support auto-upgrade of the data services extension. Recently some backend changes were made on the Arc enabled kubernetes side that requires us to set additional parameters to get this desired effect of disabling the auto-upgrade of Arc data services extension. In light of this change and to keep the deployment experience simple for our customers, we decided to remove the individual experience and keep only the "all-in-one" experience. This is much simpler and easier to use and sets all the parameters correctly behind the scenes. Without setting all these parameters correctly, there is a chance of the bootstrapper getting automatically updated while the Data controller remaining at a certain version causing a drift in the versions potentially causing issues when you upgrade. So, going forward, the recommendation is to use the simpler experience with a predictable behavior.
  

 

Stay tuned for more updates in the next release. Feel free to ask if you have any questions in the comments section below.

 

Continue reading...