Merge pull request #17402 from s1monw/issues/17280
Remove ability to specify arbitrary node attributes with `node.` prefix Today the basic node settings like `node.data` and `node.master` can't really be fully validated since we allow to specify custom user attributes on the node level. We have to, in order to support that, add a wildcard setting for `node.*` to let these setting pass validation. Instead we should require a more contraint prefix like `node.attr.` that defines a namespace that is reserved for user attributes. This commit adds a new namespace for attributes in `node.attr`. Closes #17280
This commit is contained in:
Simon Willnauer
commit
af976a6673
|
|
@ -258,7 +258,7 @@ class ClusterFormationTasks {
|
|||
'path.repo' : "${node.sharedDir}/repo",
|
||||
'path.shared_data' : "${node.sharedDir}/",
|
||||
// Define a node attribute so we can test that it exists
|
||||
'node.testattr' : 'test',
|
||||
'node.attr.testattr' : 'test',
|
||||
'repositories.url.allowed_urls': 'http://snapshot.test*'
|
||||
]
|
||||
esConfig['http.port'] = node.config.httpPort
|
||||
|
|
|
|||
|
|
@ -68,20 +68,17 @@ public class DiscoveryNodeService extends AbstractComponent {
|
|||
public DiscoveryNode buildLocalNode(TransportAddress publishAddress) {
|
||||
final String nodeId = generateNodeId(settings);
|
||||
Map<String, String> attributes = new HashMap<>(Node.NODE_ATTRIBUTES.get(this.settings).getAsMap());
|
||||
if (attributes.containsKey("client")) {
|
||||
throw new IllegalArgumentException("node.client setting is no longer supported, use " + Node.NODE_MASTER_SETTING.getKey()
|
||||
+ ", " + Node.NODE_DATA_SETTING.getKey() + " and " + Node.NODE_INGEST_SETTING.getKey() + " explicitly instead");
|
||||
Set<DiscoveryNode.Role> roles = new HashSet<>();
|
||||
if (Node.NODE_INGEST_SETTING.get(settings)) {
|
||||
roles.add(DiscoveryNode.Role.INGEST);
|
||||
}
|
||||
if (Node.NODE_MASTER_SETTING.get(settings)) {
|
||||
roles.add(DiscoveryNode.Role.MASTER);
|
||||
}
|
||||
if (Node.NODE_DATA_SETTING.get(settings)) {
|
||||
roles.add(DiscoveryNode.Role.DATA);
|
||||
}
|
||||
|
||||
attributes.remove("name"); // name is extracted in other places
|
||||
Set<DiscoveryNode.Role> roles = new HashSet<>();
|
||||
for (DiscoveryNode.Role role : DiscoveryNode.Role.values()) {
|
||||
String isRoleEnabled = attributes.remove(role.getRoleName());
|
||||
//all existing roles default to true
|
||||
if (isRoleEnabled == null || Booleans.parseBooleanExact(isRoleEnabled)) {
|
||||
roles.add(role);
|
||||
}
|
||||
}
|
||||
for (CustomAttributesProvider provider : customAttributesProviders) {
|
||||
try {
|
||||
Map<String, String> customAttributes = provider.buildAttributes();
|
||||
|
|
|
|||
|
|
@ -143,9 +143,7 @@ public class Node implements Closeable {
|
|||
public static final Setting<Boolean> NODE_INGEST_SETTING =
|
||||
Setting.boolSetting("node.ingest", true, Property.NodeScope);
|
||||
public static final Setting<String> NODE_NAME_SETTING = Setting.simpleString("node.name", Property.NodeScope);
|
||||
// this sucks that folks can mistype data, master or ingest and get away with it.
|
||||
// TODO: we should move this to node.attribute.${name} = ${value} instead.
|
||||
public static final Setting<Settings> NODE_ATTRIBUTES = Setting.groupSetting("node.", Property.NodeScope);
|
||||
public static final Setting<Settings> NODE_ATTRIBUTES = Setting.groupSetting("node.attr.", Property.NodeScope);
|
||||
|
||||
|
||||
private static final String CLIENT_TYPE = "node";
|
||||
|
|
|
|||
|
|
@ -61,7 +61,7 @@ public class AwarenessAllocationIT extends ESIntegTestCase {
|
|||
|
||||
|
||||
logger.info("--> starting 2 nodes on the same rack");
|
||||
internalCluster().startNodesAsync(2, Settings.settingsBuilder().put(commonSettings).put("node.rack_id", "rack_1").build()).get();
|
||||
internalCluster().startNodesAsync(2, Settings.settingsBuilder().put(commonSettings).put("node.attr.rack_id", "rack_1").build()).get();
|
||||
|
||||
createIndex("test1");
|
||||
createIndex("test2");
|
||||
|
|
@ -74,7 +74,7 @@ public class AwarenessAllocationIT extends ESIntegTestCase {
|
|||
ensureGreen();
|
||||
|
||||
logger.info("--> starting 1 node on a different rack");
|
||||
final String node3 = internalCluster().startNode(Settings.settingsBuilder().put(commonSettings).put("node.rack_id", "rack_2").build());
|
||||
final String node3 = internalCluster().startNode(Settings.settingsBuilder().put(commonSettings).put("node.attr.rack_id", "rack_2").build());
|
||||
|
||||
// On slow machines the initial relocation might be delayed
|
||||
assertThat(awaitBusy(
|
||||
|
|
@ -113,10 +113,10 @@ public class AwarenessAllocationIT extends ESIntegTestCase {
|
|||
|
||||
logger.info("--> starting 4 nodes on different zones");
|
||||
List<String> nodes = internalCluster().startNodesAsync(
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.zone", "a").build(),
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.zone", "b").build(),
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.zone", "b").build(),
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.zone", "a").build()
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.attr.zone", "a").build(),
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.attr.zone", "b").build(),
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.attr.zone", "b").build(),
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.attr.zone", "a").build()
|
||||
).get();
|
||||
String A_0 = nodes.get(0);
|
||||
String B_0 = nodes.get(1);
|
||||
|
|
@ -159,8 +159,8 @@ public class AwarenessAllocationIT extends ESIntegTestCase {
|
|||
|
||||
logger.info("--> starting 2 nodes on zones 'a' & 'b'");
|
||||
List<String> nodes = internalCluster().startNodesAsync(
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.zone", "a").build(),
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.zone", "b").build()
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.attr.zone", "a").build(),
|
||||
Settings.settingsBuilder().put(commonSettings).put("node.attr.zone", "b").build()
|
||||
).get();
|
||||
String A_0 = nodes.get(0);
|
||||
String B_0 = nodes.get(1);
|
||||
|
|
@ -183,7 +183,7 @@ public class AwarenessAllocationIT extends ESIntegTestCase {
|
|||
assertThat(counts.get(B_0), equalTo(5));
|
||||
logger.info("--> starting another node in zone 'b'");
|
||||
|
||||
String B_1 = internalCluster().startNode(Settings.settingsBuilder().put(commonSettings).put("node.zone", "b").build());
|
||||
String B_1 = internalCluster().startNode(Settings.settingsBuilder().put(commonSettings).put("node.attr.zone", "b").build());
|
||||
health = client().admin().cluster().prepareHealth().setWaitForEvents(Priority.LANGUID).setWaitForGreenStatus().setWaitForNodes("3").execute().actionGet();
|
||||
assertThat(health.isTimedOut(), equalTo(false));
|
||||
client().admin().cluster().prepareReroute().get();
|
||||
|
|
|
|||
|
|
@ -33,23 +33,12 @@ import static org.hamcrest.CoreMatchers.equalTo;
|
|||
|
||||
public class DiscoveryNodeServiceTests extends ESTestCase {
|
||||
|
||||
public void testClientNodeSettingIsProhibited() {
|
||||
Settings settings = Settings.builder().put("node.client", randomBoolean()).build();
|
||||
try {
|
||||
new DiscoveryNodeService(settings, Version.CURRENT).buildLocalNode(DummyTransportAddress.INSTANCE);
|
||||
fail("build attributes should have failed");
|
||||
} catch(IllegalArgumentException e) {
|
||||
assertThat(e.getMessage(), equalTo("node.client setting is no longer supported, use node.master, " +
|
||||
"node.data and node.ingest explicitly instead"));
|
||||
}
|
||||
}
|
||||
|
||||
public void testBuildLocalNode() {
|
||||
Map<String, String> expectedAttributes = new HashMap<>();
|
||||
int numCustomSettings = randomIntBetween(0, 5);
|
||||
Settings.Builder builder = Settings.builder();
|
||||
for (int i = 0; i < numCustomSettings; i++) {
|
||||
builder.put("node.attr" + i, "value" + i);
|
||||
builder.put("node.attr.attr" + i, "value" + i);
|
||||
expectedAttributes.put("attr" + i, "value" + i);
|
||||
}
|
||||
Set<DiscoveryNode.Role> selectedRoles = new HashSet<>();
|
||||
|
|
@ -76,7 +65,7 @@ public class DiscoveryNodeServiceTests extends ESTestCase {
|
|||
int numCustomSettings = randomIntBetween(0, 5);
|
||||
Settings.Builder builder = Settings.builder();
|
||||
for (int i = 0; i < numCustomSettings; i++) {
|
||||
builder.put("node.attr" + i, "value" + i);
|
||||
builder.put("node.attr.attr" + i, "value" + i);
|
||||
expectedAttributes.put("attr" + i, "value" + i);
|
||||
}
|
||||
DiscoveryNodeService discoveryNodeService = new DiscoveryNodeService(builder.build(), Version.CURRENT);
|
||||
|
|
|
|||
|
|
@ -46,9 +46,9 @@ public class ClusterSearchShardsIT extends ESIntegTestCase {
|
|||
protected Settings nodeSettings(int nodeOrdinal) {
|
||||
switch(nodeOrdinal) {
|
||||
case 1:
|
||||
return settingsBuilder().put(super.nodeSettings(nodeOrdinal)).put("node.tag", "B").build();
|
||||
return settingsBuilder().put(super.nodeSettings(nodeOrdinal)).put("node.attr.tag", "B").build();
|
||||
case 0:
|
||||
return settingsBuilder().put(super.nodeSettings(nodeOrdinal)).put("node.tag", "A").build();
|
||||
return settingsBuilder().put(super.nodeSettings(nodeOrdinal)).put("node.attr.tag", "A").build();
|
||||
}
|
||||
return super.nodeSettings(nodeOrdinal);
|
||||
}
|
||||
|
|
|
|||
|
|
@ -703,8 +703,8 @@ public class IndexWithShadowReplicasIT extends ESIntegTestCase {
|
|||
public void testIndexOnSharedFSRecoversToAnyNode() throws Exception {
|
||||
Path dataPath = createTempDir();
|
||||
Settings nodeSettings = nodeSettings(dataPath);
|
||||
Settings fooSettings = Settings.builder().put(nodeSettings).put("node.affinity", "foo").build();
|
||||
Settings barSettings = Settings.builder().put(nodeSettings).put("node.affinity", "bar").build();
|
||||
Settings fooSettings = Settings.builder().put(nodeSettings).put("node.attr.affinity", "foo").build();
|
||||
Settings barSettings = Settings.builder().put(nodeSettings).put("node.attr.affinity", "bar").build();
|
||||
|
||||
final InternalTestCluster.Async<List<String>> fooNodes = internalCluster().startNodesAsync(2, fooSettings);
|
||||
final InternalTestCluster.Async<List<String>> barNodes = internalCluster().startNodesAsync(2, barSettings);
|
||||
|
|
|
|||
|
|
@ -535,8 +535,8 @@ public class IndexRecoveryIT extends ESIntegTestCase {
|
|||
// start a master node
|
||||
internalCluster().startNode(nodeSettings);
|
||||
|
||||
InternalTestCluster.Async<String> blueFuture = internalCluster().startNodeAsync(Settings.builder().put("node.color", "blue").put(nodeSettings).build());
|
||||
InternalTestCluster.Async<String> redFuture = internalCluster().startNodeAsync(Settings.builder().put("node.color", "red").put(nodeSettings).build());
|
||||
InternalTestCluster.Async<String> blueFuture = internalCluster().startNodeAsync(Settings.builder().put("node.attr.color", "blue").put(nodeSettings).build());
|
||||
InternalTestCluster.Async<String> redFuture = internalCluster().startNodeAsync(Settings.builder().put("node.attr.color", "red").put(nodeSettings).build());
|
||||
final String blueNodeName = blueFuture.get();
|
||||
final String redNodeName = redFuture.get();
|
||||
|
||||
|
|
|
|||
|
|
@ -14,7 +14,7 @@ attribute as follows:
|
|||
|
||||
[source,sh]
|
||||
------------------------
|
||||
bin/elasticsearch -Ees.node.rack=rack1 -Ees.node.size=big <1>
|
||||
bin/elasticsearch -Ees.node.attr.rack=rack1 -Ees.node.attr.size=big <1>
|
||||
------------------------
|
||||
<1> These attribute settings can also be specified in the `elasticsearch.yml` config file.
|
||||
|
||||
|
|
|
|||
|
|
@ -26,6 +26,13 @@ should be used instead.
|
|||
The `name` setting has been removed and is replaced by `node.name`. Usage of
|
||||
`-Dname=some_node_name` is not supported anymore.
|
||||
|
||||
==== Node attribute settings
|
||||
|
||||
Node level attributes used for allocation filtering, forced awareness or other node identification / grouping
|
||||
must be prefixed with `node.attr`. In previous versions it was possible to specify node attributes with the `node.`
|
||||
prefix. All node attributes except of `node.master`, `node.data` and `node.ingest` must be moved to the new `node.attr.`
|
||||
namespace.
|
||||
|
||||
==== Node types settings
|
||||
|
||||
The `node.client` setting has been removed. A node with such a setting set will not
|
||||
|
|
|
|||
|
|
@ -21,7 +21,7 @@ attribute called `rack_id` -- we could use any attribute name. For example:
|
|||
|
||||
[source,sh]
|
||||
----------------------
|
||||
./bin/elasticsearch -Ees.node.rack_id=rack_one <1>
|
||||
./bin/elasticsearch -Ees.node.attr.rack_id=rack_one <1>
|
||||
----------------------
|
||||
<1> This setting could also be specified in the `elasticsearch.yml` config file.
|
||||
|
||||
|
|
@ -37,12 +37,12 @@ For our example, we'll set the value in the config file:
|
|||
cluster.routing.allocation.awareness.attributes: rack_id
|
||||
--------------------------------------------------------
|
||||
|
||||
With this config in place, let's say we start two nodes with `node.rack_id`
|
||||
With this config in place, let's say we start two nodes with `node.attr.rack_id`
|
||||
set to `rack_one`, and we create an index with 5 primary shards and 1 replica
|
||||
of each primary. All primaries and replicas are allocated across the two
|
||||
nodes.
|
||||
|
||||
Now, if we start two more nodes with `node.rack_id` set to `rack_two`,
|
||||
Now, if we start two more nodes with `node.attr.rack_id` set to `rack_two`,
|
||||
Elasticsearch will move shards across to the new nodes, ensuring (if possible)
|
||||
that no two copies of the same shard will be in the same rack. However if `rack_two`
|
||||
were to fail, taking down both of its nodes, Elasticsearch will still allocate the lost
|
||||
|
|
@ -102,10 +102,10 @@ cluster.routing.allocation.awareness.attributes: zone
|
|||
-------------------------------------------------------------------
|
||||
<1> We must list all possible values that the `zone` attribute can have.
|
||||
|
||||
Now, if we start 2 nodes with `node.zone` set to `zone1` and create an index
|
||||
Now, if we start 2 nodes with `node.attr.zone` set to `zone1` and create an index
|
||||
with 5 shards and 1 replica. The index will be created, but only the 5 primary
|
||||
shards will be allocated (with no replicas). Only when we start more shards
|
||||
with `node.zone` set to `zone2` will the replicas be allocated.
|
||||
with `node.attr.zone` set to `zone2` will the replicas be allocated.
|
||||
|
||||
The `cluster.routing.allocation.awareness.*` settings can all be updated
|
||||
dynamically on a live cluster with the
|
||||
|
|
|
|||
Loading…
Reference in New Issue