use of org.elasticsearch.test.discovery.TestZenDiscovery in project elasticsearch by elastic.
the class DiscoveryWithServiceDisruptionsIT method testUnicastSinglePingResponseContainsMaster.
/**
* A 4 node cluster with m_m_n set to 3 and each node has one unicast endpoint. One node partitions from the master node.
* The temporal unicast responses is empty. When partition is solved the one ping response contains a master node.
* The rejoining node should take this master node and connect.
*/
public void testUnicastSinglePingResponseContainsMaster() throws Exception {
List<String> nodes = startCluster(4, -1, new int[] { 0 });
// Figure out what is the elected master node
final String masterNode = internalCluster().getMasterName();
logger.info("---> legit elected master node={}", masterNode);
List<String> otherNodes = new ArrayList<>(nodes);
otherNodes.remove(masterNode);
// <-- Don't isolate the node that is in the unicast endpoint for all the other nodes.
otherNodes.remove(nodes.get(0));
final String isolatedNode = otherNodes.get(0);
// Forcefully clean temporal response lists on all nodes. Otherwise the node in the unicast host list
// includes all the other nodes that have pinged it and the issue doesn't manifest
ZenPing zenPing = ((TestZenDiscovery) internalCluster().getInstance(Discovery.class)).getZenPing();
if (zenPing instanceof UnicastZenPing) {
((UnicastZenPing) zenPing).clearTemporalResponses();
}
// Simulate a network issue between the unlucky node and elected master node in both directions.
NetworkDisruption networkDisconnect = new NetworkDisruption(new TwoPartitions(masterNode, isolatedNode), new NetworkDisconnect());
setDisruptionScheme(networkDisconnect);
networkDisconnect.startDisrupting();
// Wait until elected master has removed that the unlucky node...
ensureStableCluster(3, masterNode);
// The isolate master node must report no master, so it starts with pinging
assertNoMaster(isolatedNode);
networkDisconnect.stopDisrupting();
// Wait until the master node sees all 4 nodes again.
ensureStableCluster(4);
// The elected master shouldn't have changed, since the isolated node never could have elected himself as
// master since m_m_n of 3 could never be satisfied.
assertMaster(masterNode, nodes);
}
use of org.elasticsearch.test.discovery.TestZenDiscovery in project elasticsearch by elastic.
the class DiscoveryWithServiceDisruptionsIT method startCluster.
private List<String> startCluster(int numberOfNodes, int minimumMasterNode, @Nullable int[] unicastHostsOrdinals) throws ExecutionException, InterruptedException {
configureCluster(numberOfNodes, unicastHostsOrdinals, minimumMasterNode);
List<String> nodes = internalCluster().startNodes(numberOfNodes);
ensureStableCluster(numberOfNodes);
// TODO: this is a temporary solution so that nodes will not base their reaction to a partition based on previous successful results
ZenPing zenPing = ((TestZenDiscovery) internalCluster().getInstance(Discovery.class)).getZenPing();
if (zenPing instanceof UnicastZenPing) {
((UnicastZenPing) zenPing).clearTemporalResponses();
}
return nodes;
}
use of org.elasticsearch.test.discovery.TestZenDiscovery in project elasticsearch by elastic.
the class DiscoveryWithServiceDisruptionsIT method testIsolatedUnicastNodes.
public void testIsolatedUnicastNodes() throws Exception {
List<String> nodes = startCluster(4, -1, new int[] { 0 });
// Figure out what is the elected master node
final String unicastTarget = nodes.get(0);
Set<String> unicastTargetSide = new HashSet<>();
unicastTargetSide.add(unicastTarget);
Set<String> restOfClusterSide = new HashSet<>();
restOfClusterSide.addAll(nodes);
restOfClusterSide.remove(unicastTarget);
// Forcefully clean temporal response lists on all nodes. Otherwise the node in the unicast host list
// includes all the other nodes that have pinged it and the issue doesn't manifest
ZenPing zenPing = ((TestZenDiscovery) internalCluster().getInstance(Discovery.class)).getZenPing();
if (zenPing instanceof UnicastZenPing) {
((UnicastZenPing) zenPing).clearTemporalResponses();
}
// Simulate a network issue between the unicast target node and the rest of the cluster
NetworkDisruption networkDisconnect = new NetworkDisruption(new TwoPartitions(unicastTargetSide, restOfClusterSide), new NetworkDisconnect());
setDisruptionScheme(networkDisconnect);
networkDisconnect.startDisrupting();
// Wait until elected master has removed that the unlucky node...
ensureStableCluster(3, nodes.get(1));
// The isolate master node must report no master, so it starts with pinging
assertNoMaster(unicastTarget);
networkDisconnect.stopDisrupting();
// Wait until the master node sees all 3 nodes again.
ensureStableCluster(4);
}
Aggregations