Low level understanding of How ReadFrom.ANY Works?

[harshsunil1]

I've been using the Redis Lettuce Client with the connection setting configured as follows:

connection.setReadFrom(ReadFrom.LOWEST_LATENCY);

When monitoring the ElastiCache metrics, I noticed that the traffic pattern and CPU usage were not balanced, which led me to change the setting to:

connection.setReadFrom(ReadFrom.ANY);

With this new setting, I observed a more even load distribution, but this also resulted in higher CPU utilization. Could someone help me understand how the load is distributed with these configurations? Specifically:

How is the load distributed between the master and replica nodes when using ANY?
How frequently are the connections refreshed or rebalanced between nodes?
What could be the cause of the increased CPU utilization after switching to ReadFrom.ANY?
Your insights would be greatly appreciated.

[tishun]

Hey @harshsunil1 ,

You can read more in the manual on ReadFrom here, specifically:

Setting	Description
MASTER	Default mode. Read from the current master node.
MASTER_PREFERRED	Read from the master, but if it is unavailable, read from replica nodes.
REPLICA	Read from replica nodes.
REPLICA_PREFERRED	Read from the replica nodes, but if none is unavailable, read from the master.
LOWEST_LATENCY	Read from any node of the cluster with the lowest latency.
ANY	Read from any node of the cluster.
ANY_REPLICA	Read from any replica of the cluster.

How is the load distributed between the master and replica nodes when using ANY?

Every time a read is done a random node is selected.

How frequently are the connections refreshed or rebalanced between nodes?

Not sure what "refreshed" means in this case. There is no balancing done.

What could be the cause of the increased CPU utilization after switching to ReadFrom.ANY?

Can't see why, but you need to talk to Amazon about that.

Important

All these settings affect the client side / driver. The different providers might allow different vendor-specific ways to balance the load (either through DNS or some other approach) and we can't help with that.

[harshsunil1]

Not sure what "refreshed" means in this case. There is no balancing done.

If we set ReadFrom to LOWEST_LATENCY - nodes are determined based on cluster topology refresh. In my snippet i have configured it to an hour.
Now for ANY - The requests are evenly distributed, does it make connection with both the shard and it's replica or periodically keep on refreshing the connection with the nodes?
The document which you have shared only shares I level details and i want to understand the low level details to understand what can increase the CPU.