Question about usage of kubernetes headless service and aerospike java client

Client Libraries Java Client
,
1

Hi, I have a question regarding the client behaviour when I pass in the hostname of a kubernetes headless service.

I am using java client version 5.1.11

Since the headless service should resolve to all the backing pod ip addresses, does that mean that the java aerospike client will create multiple connections, one to each aerospike pod?

Also, (this was a more general question), but in the case of a connection failure to one ip, does the aerospike java client do any fallback to the other ips in the pool? Or is the pooling only used in the initial connection step (ie connect to any ip in the pool, then forget the rest of the ips).

Any info is appreciated, thanks!

2

The java client resolves the kubernetes hostname to an IP address and then connects with that server node. The client then asks that server node for the IP addresses of it’s peers. The server can be configured with a specific “access-address” when the default IP address is not accessible by the client.

The client then connects with each peer node using the returned IP addresses. The client also creates a connection pool for each node. The client connects to the first valid IP address and then forgets the rest of the ips.