Does Dolt have a high availability solution?

[lihh1992]

Our Dolt deployment architecture is as follows: We use the dolt-sql-server 1.79.1 image, deployed as a StatefulSet in Kubernetes. In this setup, dolt-0 serves as the primary node, and dolt-1 serves as the replica node. dolt-1 syncs data from the dolt-0 node. Applications connect directly to the dolt-0 node.

Now, suppose the dolt-0 node becomes unavailable. How can we automatically detect the failure of the primary node and automatically switch to the replica node without manual intervention, thereby achieving high availability for the database and applications?

[timsehn]

This is a question for @reltuk or @bheni, both of which are out until the new year.

In Hosted Dolt, we run an external daemon running health checks and fail over after a certain number of checks.

[bpf120]

@lihh1992 , thanks for all of the issues you have filed. We'd love to learn more about your Dolt use case. Feel free to email me or swing by our Discord if you'd like to share.

[lihh1992]

This is a question for @reltuk or @bheni, both of which are out until the new year.

In Hosted Dolt, we run an external daemon running health checks and fail over after a certain number of checks.

May I ask, when is the approximate release time for this update?

[reltuk]

@lihh1992 We do not currently have plans to make the hosted control plane logic available as open source.

If the question is how long does hosted take to detect the failure and promote a new primary, @bheni is the correct person to answer it. But, it will depend on the cloud platform and the network topology being used for the hosted instance, since sometimes promoting a new primary involves changing cloud configuration for things like target groups in a cloud load balancer. AFAIK, the current approaches taken by hosted and by other Dolt customers in the context of things like Kubernetes take on the order of a few minutes to perform a failover. So they are not very high-availability for high throughput use cases, but they can help to stop the pain and recover automatically after a failure of the primary.