Ping on first usage
editPing on first usage
editPinging is enabled by default for the Static, Sniffing and Sticky connection pools. This means that the first time a node is used or resurrected, a ping is issued a with a small (configurable) timeout, allowing the client to fail and fallover to a healthy node much faster than attempting a request, that may be heavier than a ping.
Here’s an example with a cluster with two nodes where the second node fails on ping
var audit = new Auditor(() => VirtualClusterWith .Nodes(2) .Ping(p => p.Succeeds(Always)) .Ping(p => p.OnPort(9201).FailAlways()) .ClientCalls(c=>c.SucceedAlways()) .StaticConnectionPool() .AllDefaults() );
When making the calls, the first call goes to 9200 which succeeds, and the 2nd call does a ping on 9201 because it’s used for the first time. The ping fails so we wrap over to node 9200 which we’ve already pinged.
Finally we assert that the connectionpool has one node that is marked as dead
await audit.TraceCalls( new ClientCall { { PingSuccess, 9200}, { HealthyResponse, 9200}, { pool => { pool.Nodes.Where(n=>!n.IsAlive).Should().HaveCount(0); } } }, new ClientCall { { PingFailure, 9201}, { HealthyResponse, 9200}, { pool => pool.Nodes.Where(n=>!n.IsAlive).Should().HaveCount(1) } } );
A cluster with 4 nodes where the second and third pings fail
var audit = new Auditor(() => VirtualClusterWith .Nodes(4) .Ping(p => p.SucceedAlways()) .Ping(p => p.OnPort(9201).FailAlways()) .Ping(p => p.OnPort(9202).FailAlways()) .ClientCalls(c=>c.SucceedAlways()) .StaticConnectionPool() .AllDefaults() ); await audit.TraceCalls( new ClientCall { { PingSuccess, 9200}, { HealthyResponse, 9200}, { pool => pool.Nodes.Where(n=>!n.IsAlive).Should().HaveCount(0) } }, new ClientCall { { PingFailure, 9201}, { PingFailure, 9202}, { PingSuccess, 9203}, { HealthyResponse, 9203}, { pool => pool.Nodes.Where(n=>!n.IsAlive).Should().HaveCount(2) } } );
The first call goes to 9200, which succeeds |
|
The 2nd call does a ping on 9201 because its used for the first time. This fails |
|
So we ping 9202. This also fails |
|
We then ping 9203 because we haven’t used it before and it succeeds |
|
Finally, we assert that the connection pool has two nodes that are marked as dead |
All nodes are pinged on first use, provided they are healthy
var audit = new Auditor(() => VirtualClusterWith .Nodes(4) .Ping(p => p.SucceedAlways()) .ClientCalls(c=>c.SucceedAlways()) .StaticConnectionPool() .AllDefaults() ); await audit.TraceCalls( new ClientCall { { PingSuccess, 9200}, { HealthyResponse, 9200} }, new ClientCall { { PingSuccess, 9201}, { HealthyResponse, 9201} }, new ClientCall { { PingSuccess, 9202}, { HealthyResponse, 9202} }, new ClientCall { { PingSuccess, 9203}, { HealthyResponse, 9203} }, new ClientCall { { HealthyResponse, 9200} }, new ClientCall { { HealthyResponse, 9201} }, new ClientCall { { HealthyResponse, 9202} }, new ClientCall { { HealthyResponse, 9203} }, new ClientCall { { HealthyResponse, 9200} } );