honeymoose
/
OpenSearch
mirror of https://github.com/honeymoose/OpenSearch.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Simplify InternalTestCluster.fullRestart (#50218) 

With node ordinals gone, there's no longer a need for such a complicated full cluster restart
procedure (as we can now uniquely associate nodes to data folders).

Follow-up to #41652
This commit is contained in:
Yannick Welsch 2019-12-17 13:51:51 +01:00
parent 098f540f9d
commit 1f981580aa
1 changed files with 9 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
test/framework/src/main/java/org/elasticsearch/test/InternalTestCluster.java
View File

@ -55,6 +55,7 @@ import org.elasticsearch.cluster.routing.allocation.DiskThresholdSettings;
import org.elasticsearch.cluster.routing.allocation.decider.ThrottlingAllocationDecider;
import org.elasticsearch.cluster.service.ClusterService;
import org.elasticsearch.common.Nullable;
import org.elasticsearch.common.Randomness;
import org.elasticsearch.common.Strings;
import org.elasticsearch.common.breaker.CircuitBreaker;
import org.elasticsearch.common.component.LifecycleListener;
@ -1895,51 +1896,30 @@ public final class InternalTestCluster extends TestCluster {
public synchronized void fullRestart(RestartCallback callback) throws Exception {
int numNodesRestarted = 0;
final Settings[] newNodeSettings = new Settings[nextNodeId.get()];
Map<Set<DiscoveryNodeRole>, List<NodeAndClient>> nodesByRoles = new HashMap<>();
Set[] rolesOrderedByOriginalStartupOrder = new Set[nextNodeId.get()];
final int minMasterNodes = autoManageMasterNodes ? getMinMasterNodes(getMasterNodesCount()) : -1;
final int nodeCount = nodes.size();
final List<NodeAndClient> toStartAndPublish = new ArrayList<>(); // we want to start nodes in one go
for (NodeAndClient nodeAndClient : nodes.values()) {
callback.doAfterNodes(numNodesRestarted++, nodeAndClient.nodeClient());
logger.info("Stopping and resetting node [{}] ", nodeAndClient.name);
if (activeDisruptionScheme != null) {
activeDisruptionScheme.removeFromNode(nodeAndClient.name, this);
}
DiscoveryNode discoveryNode = getInstanceFromNode(ClusterService.class, nodeAndClient.node()).localNode();
final Settings newSettings = nodeAndClient.closeForRestart(callback, minMasterNodes);
newNodeSettings[nodeAndClient.nodeAndClientId()] = newSettings;
rolesOrderedByOriginalStartupOrder[nodeAndClient.nodeAndClientId()] = discoveryNode.getRoles();
nodesByRoles.computeIfAbsent(discoveryNode.getRoles(), k -> new ArrayList<>()).add(nodeAndClient);
toStartAndPublish.add(nodeAndClient);
}
callback.onAllNodesStopped();
assert nodesByRoles.values().stream().mapToInt(List::size).sum() == nodeCount;
// randomize start up order
Randomness.shuffle(toStartAndPublish);
// randomize start up order, but making sure that:
// 1) A data folder that was assigned to a data node will stay so
// 2) Data nodes will get the same node lock ordinal range, so custom index paths (where the ordinal is used)
// will still belong to data nodes
for (List<NodeAndClient> sameRoleNodes : nodesByRoles.values()) {
Collections.shuffle(sameRoleNodes, random);
}
final List<NodeAndClient> startUpOrder = new ArrayList<>();
for (Set roles : rolesOrderedByOriginalStartupOrder) {
if (roles == null) {
// if some nodes were stopped, we want have a role for that ordinal
continue;
}
final List<NodeAndClient> nodesByRole = nodesByRoles.get(roles);
startUpOrder.add(nodesByRole.remove(0));
}
assert nodesByRoles.values().stream().mapToInt(List::size).sum() == 0;
for (NodeAndClient nodeAndClient : startUpOrder) {
logger.info("creating node [{}] ", nodeAndClient.name);
nodeAndClient.recreateNode(newNodeSettings[nodeAndClient.nodeAndClientId()], () -> rebuildUnicastHostFiles(startUpOrder));
for (NodeAndClient nodeAndClient : toStartAndPublish) {
logger.info("recreating node [{}] ", nodeAndClient.name);
nodeAndClient.recreateNode(newNodeSettings[nodeAndClient.nodeAndClientId()], () -> rebuildUnicastHostFiles(toStartAndPublish));
}
startAndPublishNodesAndClients(startUpOrder);
startAndPublishNodesAndClients(toStartAndPublish);
if (callback.validateClusterForming()) {
validateClusterFormed();