S3 cluster management

This page describes how to set up, update, monitor, and delete an S3 cluster.

Considerations

  • Performance scale: The S3 service can be exposed from the cluster servers, ranging from three servers to the entire cluster. The service performance scales linearly as the S3 cluster scales. Depending on the workload, you may need several Frontend cores to gain maximum performance.

  • Cluster-wide configuration filesystem: The S3 protocol requires a persistent cluster-wide configuration filesystem (see Set the global configuration filesystem).

Round-robin DNS or load balancer

To ensure load balancing between the S3 clients on the different WEKA servers serving S3, it is recommended to configure a round-robin DNS entry that resolves to the list of servers' IPs.

A DNS server that supports health checks can help with resiliency if any servers serving S3 become unresponsive.

Even a robust DNS server or load-balancer may become overloaded with an extreme load. You can also use a client-side load balancer, where each client checks the health of each S3 container in the cluster. An example of such a load balancer is the open-source Sidekick Load Balancer.

Related information

Round-robin DNS

Sidekick Load Balancer

Related topics

Manage the S3 service using the GUI

Manage the S3 service using the CLI

Last updated