Cluster manager handles all of the complexity of knowing which hosts are available and healthy, load balancing, thread local storage of upstream connection data (since most Envoy code is written to be single threaded), upstream connection type (TCP/IP, UDS), upstream protocol where applicable (HTTP/1.1, HTTP/2), etc.

Cluster warming

New cluster will not appear to exist, old cluster upgraded will not take places until warmed

clusters do not become available until the following operations have taken place.

• Initial service discovery load
• Initial active health check pass

Service discovery

Supported service discovery types

Static

network name (IP address/port, unix domain socket, etc.)

Strict DNS

continuously and asynchronously resolve the specified DNS targets

if the query returns three IP addresses, Envoy will assume the cluster has three hosts, and all three should be load balanced to

• the minimum of all record TTLs is 0, default is 5000ms
• dns_failure_refresh_rate is for failure, if not configured, use dns_refresh_rate

In strict DNS mode, envoy assume the cluster is changing during every resolution which would lead to draining connection pools, connection cycling, because dns is based on RR policy

Logical DNS

logical DNS cluster only uses the first IP address returned when a new connection needs to be initiated

When interacting with large scale web services, this is the best: asynchronous/eventually consistent DNS resolution, long lived connections, and zero blocking in the forwarding path.