Blog Post

SQL Server Support Blog
3 MIN READ

SQL pod may get stuck in "ContainerCreating" status when you stop the node instance on AKS

SiruiWang's avatar
SiruiWang
Icon for Microsoft rankMicrosoft
Dec 17, 2020
When we deploy SQL Server on AKS, sometimes we may find SQL HA is not working as expect.   For example, when we deploy AKS using our default sample with 2 nodes: https://docs.microsoft.com/en-u...
Updated Feb 19, 2021
Version 2.0
aks
azure
high availability
linux
sql
tdihp's avatar
tdihp
Icon for Microsoft rankMicrosoft
Dec 26, 2020

Thanks for clarifying this. The tutorial includes a simple configuration where only 1 pod is configured.

However even though Kubernetes/AKS can add some more reliability by having the pod re-scheduled, this is not a silver bullet for availability solution.

 

Virtual Machines can have multiple failure patterns. In some of those the PV (disk) could also be impacted, and there's a risk of the disk being stuck on the failing node. The approach assumes the disk can be successfully detached/attached to another VM during the recovery, this is not always true. Adding to that the sqlserver instance wouldn't be zone redundant.

 

I'd suggest to consider active/backup deployment pattern for a more reliable deployment. It would require considerable effort to deploy in Kubernetes but it would be a great to have. See kubedb project for an example of deploying reliable databases in Kubernetes, which demonstrates all sorts of replication/standby patterns can also be implemented in Kubernetes, postgres-ha helm chart also demonstrates a HA setting.