This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(128, 33%, 22%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EMH%3C/text%3E%3C/svg%3E)
Hello,
We are a large financial firm in the process of moving our SQL Server environment to Linux. We are planning on using Always On Availability Groups in our solution as we do today on Windows Server. We have installed and proved our this configuration with bare metal and VMs using Pacemaker cluster management. We are now looking at utilizing containers with either Pacemaker_remote or with Kubernetes, both of which we are unsure at this time if our configuration is supported.
Can anyone confirm if Microsoft is going to or does currently support Always On Availability Groups (cluster type=EXTERNAL) on Kubernetes? And if so, where can you download the Kubernetes operator?
Thank you
-Martin
So Kubernetes makes my head spin quite a bit, but it seems that Microsoft are using Kubernetes both here and there. I found this blog post from my MVP Andrew Nocentino, which might be helpful: https://www.centinosystems.com/blog/sql/deploying-sql-server-in-kubernetes/
It does help although it was posted back in 2018 as are the code samples for the Microsoft created Kubernetes operator which leads me to believe this functionality never made it out of "testing/preview". I can't find any official doc or confirmation from Microsoft, or anything recent, like within the last year of running SQL Server Availability Groups in Kubernetes. I know we can definitely run SQL Server in Kubernetes but my company is insistent that we use AGs for our HADR solution so I'm trying to determine all the viable, supported options...
In the Google search, I saw a date that was more recent, but then I did not look closely on the blog post itself. I guess the date Google showed was the date for the last comment to the post.
Anyway, reading the comments from the end, I see that Andrew has posted this: Hi there…as far as I know, AGs aren’t available in the GA release for 2019.
I'll try to ping Andrew, to see if he has anything to add.
thanks for that!! I didn't read down in the comments. I had heard it didn't make GA but as it's now been some time since then I was hoping it was now a support option. If you find anything else out please post here. and thanks again!
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(118.4, 18%, 21%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EAN%3C/text%3E%3C/svg%3E)
Hi there, the current landscape is...AGs in Kubernetes does not support automatic failover. If you need automatic failover with AGs using multiple replicas check out Azure Arc Enabled Managed Instance in Azure Arc Enabled Data Services...but that's still in preview.
Running SQL Server in Kubernetes will get you HA (failover cluster like functionality) in that if a Pod fails it will be replaced quickly. If a Node fails, your storage subsystem and the Volumes in k8s will need some tweaking to move to another node in the cluster quickly. If you're looking for DR you can add on log shipping.
Thank you Anthony!
That clears it up. I had suspected as much but held out for that small bit of hope. My client is set on using Availability Groups in an on-prem setting AND moving their SQL Server instances off Windows and on to Linux, they have several hundred instances and thousands of databases so it's going to be quite the effort. We will probably go with the most straight forward approach of using mutli-site Pacemaker clusters across their three data centers with AGs. We've proved this option out already but I was also investigating what HADR options would exist if we utilized containers with either Docker, k8s, or in-house tooling in an effort to bypass Pacemaker.
Thanks again for the information, it's much appreciated!
-Martin