Abdullah:
Hey guys, im new to networking, but from my understanding from this is that multiple connection are established with etcd, but shouldn’t it only be one? the API server?
Alistair Mackay:
Hi @Abdullah
What you’re seeing here is a pool of connections between the API server and etcd.
API server is able to send multiple requests simultaneously to etcd allowing it to process many queries/updates at the same time. On a large cluster, if API server could only process one request at a time in sequence, then it would be extremely unresponsive!
Abdullah:
You’re awesome Alistair! thank you for clearing it up
Conall Ó Gribín:
yeah, most database servers behave similarly, having a connection pool means we don’t have to:
-wait for connections to come up
-worry as much about buffer overflows when data piles up that we haven’t transmitted yet
-can use a different pipe if one fails really quickly.
-we can scale the pool size up or down dynamically (if supported and enabled in config.) - not sure if this is supported by etcd though!