WARNING: Version 5.x has passed its EOL date.
This documentation is no longer being maintained and may be removed. If you are running this version, we strongly advise you to upgrade. For the latest information, see the current release documentation.
Sniffing on connection failure
editSniffing on connection failure
editSniffing on connection is enabled by default when using a connection pool that allows reseeding. The only connection pool we ship with that allows this is the Sniffing connection pool.
This can be very handy to force a refresh of the connection pool’s known healthy nodes by asking the Elasticsearch cluster itself, and a sniff tries to get the nodes by asking each node it currently knows about, until one responds.
Here we seed our connection with 5 known nodes on ports 9200-9204, of which we think 9202, 9203, 9204 are master eligible nodes. Our virtualized cluster will throw once when doing a search on 9201. This should cause a sniff to be kicked off.
var audit = new Auditor(() => Framework.Cluster .Nodes(5) .MasterEligible(9202, 9203, 9204) .ClientCalls(r => r.SucceedAlways()) .ClientCalls(r => r.OnPort(9201).Fails(Once)) .Sniff(p => p.SucceedAlways(Framework.Cluster .Nodes(3) .MasterEligible(9200, 9202) .ClientCalls(r => r.OnPort(9201).Fails(Once)) .Sniff(s => s.SucceedAlways(Framework.Cluster .Nodes(3, 9210) .MasterEligible(9210, 9212) .ClientCalls(r => r.SucceedAlways()) .Sniff(r => r.SucceedAlways()) )) )) .SniffingConnectionPool() .Settings(s => s.DisablePing().SniffOnStartup(false)) ); audit = await audit.TraceCalls( /** */ new ClientCall { { HealthyResponse, 9200 }, { pool => pool.Nodes.Count.Should().Be(5) } }, new ClientCall { { BadResponse, 9201}, { SniffOnFail }, { SniffSuccess, 9202}, { HealthyResponse, 9200}, { pool => pool.Nodes.Count.Should().Be(3) } }, new ClientCall { { BadResponse, 9201}, { SniffOnFail }, { SniffSuccess, 9200}, { HealthyResponse, 9210}, { pool => pool.Nodes.Count.Should().Be(3) } }, new ClientCall { { HealthyResponse, 9211 } }, new ClientCall { { HealthyResponse, 9212 } }, new ClientCall { { HealthyResponse, 9210 } }, new ClientCall { { HealthyResponse, 9211 } }, new ClientCall { { HealthyResponse, 9212 } }, new ClientCall { { HealthyResponse, 9210 } }, new ClientCall { { HealthyResponse, 9211 } }, new ClientCall { { HealthyResponse, 9212 } }, new ClientCall { { HealthyResponse, 9210 } } );
When the call fails on 9201, the following sniff succeeds and returns a new cluster state of healthy nodes. This cluster only has 3 nodes and the known masters are 9200 and 9202. A search on 9201 is setup to still fail once |
|
After this second failure on 9201, another sniff will happen which returns a cluster state that no longer fails but looks completely different; It’s now three nodes on ports 9210 - 9212, with 9210 and 9212 being master eligible. |
|
We assert we do a sniff on our first known master node 9202 after the failed call on 9201 |
|
Our pool should now have three nodes |
|
We assert we do a sniff on the first master node in our updated cluster |
Sniffing after ping failure
editHere we set up our cluster exactly the same as the previous setup
Only we enable pinging (default is true
) and make the ping fail
var audit = new Auditor(() => Framework.Cluster .Nodes(5) .MasterEligible(9202, 9203, 9204) .Ping(r => r.OnPort(9201).Fails(Once)) .Sniff(p => p.SucceedAlways(Framework.Cluster .Nodes(3) .MasterEligible(9200, 9202) .Ping(r => r.OnPort(9201).Fails(Once)) .Sniff(s => s.SucceedAlways(Framework.Cluster .Nodes(3, 9210) .MasterEligible(9210, 9211) .Ping(r => r.SucceedAlways()) .Sniff(r => r.SucceedAlways()) )) )) .SniffingConnectionPool() .Settings(s => s.SniffOnStartup(false)) ); audit = await audit.TraceCalls( new ClientCall { { PingSuccess, 9200 }, { HealthyResponse, 9200 }, { pool => pool.Nodes.Count.Should().Be(5) } }, new ClientCall { { PingFailure, 9201}, { SniffOnFail }, { SniffSuccess, 9202}, { PingSuccess, 9200}, { HealthyResponse, 9200}, { pool => pool.Nodes.Count.Should().Be(3) } }, new ClientCall { { PingFailure, 9201}, { SniffOnFail }, { SniffSuccess, 9200}, { PingSuccess, 9210}, { HealthyResponse, 9210}, { pool => pool.Nodes.Count.Should().Be(3) } }, new ClientCall { { PingSuccess, 9211 }, { HealthyResponse, 9211 } }, new ClientCall { { PingSuccess, 9212 }, { HealthyResponse, 9212 } }, new ClientCall { { HealthyResponse, 9210 } }, new ClientCall { { HealthyResponse, 9211 } }, new ClientCall { { HealthyResponse, 9212 } }, new ClientCall { { HealthyResponse, 9210 } } );
Client uses publish address
editvar audit = new Auditor(() => Framework.Cluster .Nodes(2) .MasterEligible(9200) .Ping(r => r.OnPort(9200).Fails(Once)) .Sniff(p => p.SucceedAlways(Framework.Cluster .Nodes(10) .MasterEligible(9200, 9202, 9201) .PublishAddress("10.0.12.1") )) .SniffingConnectionPool() .Settings(s => s.SniffOnStartup(false)) ); Action<Audit, string, int> hostAssert = (a, host, expectedPort) => { a.Node.Uri.Host.Should().Be(host); a.Node.Uri.Port.Should().Be(expectedPort); }; audit = await audit.TraceCalls( new ClientCall { { PingFailure, a => hostAssert(a, "localhost", 9200)}, { SniffOnFail }, { SniffSuccess, a => hostAssert(a, "localhost", 9200)}, { PingSuccess, a => hostAssert(a, "10.0.12.1", 9200)}, { HealthyResponse, a => hostAssert(a, "10.0.12.1", 9200)}, { pool => pool.Nodes.Count.Should().Be(10) } } );