Hi,
I want to setup a multinode timescale db with high availability. In case of any failure at any level of node, my system should not be affected since it will be running in production.
From what I understand from docs, we can have replicas for data nodes which can automatically be used in case of data node failure.
But how can we ensure this high availability for access node? In case of failure at access node a standby node should instantly replace failed node without any data loss or affecting the system.
Also is there a way to declare a node as failed in case of high latency or increased no of failed requests?
Just a gentle reminder!
Hai,
You can setup a physical streaming replica of the Access Node. Ideally, it should be a sync replica so that all the metadata changes reach the replica immediately. One can then fail over to the replica and promote it in case of issues with the primary.
For the data nodes, it’s recommended to use native replication
To declare a node as failed, one can use the alter_data_node API to mark a datanode as unavailable.
Can you please suggest some document or library you will prefer to use for setting up physical streaming replica of access node?