Was this page helpful?
ScyllaDB Java Driver is available under the Apache v2 License. ScyllaDB Java Driver is a fork of DataStax Java Driver. See Copyright here.
Caution
You're viewing documentation for a previous version of Scylla Java Driver. Switch to the latest stable version.
session.getMetadata().getTokenMap()
used for token-aware routing or analytics clients.
immutable (must invoke again to observe changes).
advanced.metadata.token-map.enabled in the configuration (defaults to true).
Metadata#getTokenMap returns information about the tokens used for data replication. It is used internally by the driver to send requests to the optimal coordinator when token-aware routing is enabled. Another typical use case is data analytics clients, for example fetching a large range of keys in parallel by sending sub-queries to each replica.
Because token metadata can be disabled, the resulting TokenMap object is wrapped in an Optional;
to access it, you can use either a functional pattern, or more traditionally test first with
isPresent and then unwrap:
Metadata metadata = session.getMetadata();
metadata.getTokenMap().ifPresent(tokenMap -> {
// do something with the map
});
if (metadata.getTokenMap().isPresent()) {
TokenMap tokenMap = metadata.getTokenMap().get();
// do something with the map
}
TokenMap methods¶For illustration purposes, let’s consider a fictitious ring with 6 tokens, and a cluster of 3 nodes that each own two tokens:
node1
/---\
/=---+ 12+---=\
: \---/ :
| |
/-+-\ /-+-\
node3 | 10| | 2 | node2
\-+-/ \-+-/
: :
| |
/---\ /-+-\
node2 | 8 | | 4 | node3
\-+-/ \-+-/
: :
| /---\ |
\=---+ 6 +---=/
\---/
node1
The first thing you can do is retrieve all the ranges, in other words describe the ring:
Set<TokenRange> ring = tokenMap.getTokenRanges();
// Returns [Murmur3TokenRange(Murmur3Token(12), Murmur3Token(2)),
// Murmur3TokenRange(Murmur3Token(2), Murmur3Token(4)),
// Murmur3TokenRange(Murmur3Token(4), Murmur3Token(6)),
// Murmur3TokenRange(Murmur3Token(6), Murmur3Token(8)),
// Murmur3TokenRange(Murmur3Token(8), Murmur3Token(10)),
// Murmur3TokenRange(Murmur3Token(10), Murmur3Token(12))]
Note: Murmur3Token is an implementation detail. The actual class depends on the partitioner
you configured in Cassandra, but in general you don’t need to worry about that. TokenMap provides
a few utility methods to parse tokens and create new instances: parse, format, newToken and
newTokenRange.
You can also retrieve the ranges and tokens owned by a specific replica:
tokenMap.getTokenRanges(node1);
// [Murmur3TokenRange(Murmur3Token(10), Murmur3Token(12)),
// Murmur3TokenRange(Murmur3Token(4), Murmur3Token(6))]
tokenMap.getTokens(node1);
// [Murmur3Token(12)), Murmur3Token(6))]
As shown here, the node owns the ranges that end with its tokens; this is because ranges are
start-exclusive and end-inclusive: ]10, 12] and ]4, 6].
Next, you can retrieve keyspace-specific information. To illustrate this, let’s use two keyspaces with different replication settings:
// RF = 1: each range is only stored on the primary replica
CREATE KEYSPACE ks1 WITH replication = {'class': 'SimpleStrategy', 'replication_factor': 1};
// RF = 2: each range is stored on the primary replica, and replicated on the next node in the ring
CREATE KEYSPACE ks2 WITH replication = {'class': 'SimpleStrategy', 'replication_factor': 2};
getReplicas finds the nodes that have the data in a given range:
TokenRange firstRange = tokenMap.getTokenRanges().iterator().next();
// Murmur3TokenRange(Murmur3Token(12), Murmur3Token(2))
Set<Node> nodes1 = tokenMap.getReplicas(CqlIdentifier.fromCql("ks1"), firstRange);
// [node2] (only the primary replica)
Set<Node> nodes2 = tokenMap.getReplicas(CqlIdentifier.fromCql("ks2"), firstRange);
// [node2, node3] (the primary replica, and the next node on the ring)
There is a also a variant that takes a primary key, to find the replicas for a particular row. In the following example, let’s assume that the key hashes to the token “1” with the current partitioner:
String pk = "johndoe@example.com";
// You need to manually encode the key as binary:
ByteBuffer encodedPk = TypeCodecs.TEXT.encode(pk, session.getContext().getProtocolVersion());
Set<Node> nodes1 = tokenMap.getReplicas(CqlIdentifier.fromInternal("ks1"), encodedPk);
// Assuming the key hashes to "1", it is in the ]12, 2] range
// => [node2] (only the primary replica)
Set<Node> nodes2 = tokenMap.getReplicas(CqlIdentifier.fromCql("ks2"), encodedPk);
// [node2, node3] (the primary replica, and the next node on the ring)
Finally, you can go the other way, and find the token ranges that a node stores for a given keyspace:
Set<TokenRange> ranges1 = tokenMap.getTokenRanges(CqlIdentifier.fromCql("ks1"), node1);
// [Murmur3TokenRange(Murmur3Token(4), Murmur3Token(6)),
// Murmur3TokenRange(Murmur3Token(10), Murmur3Token(12))]
// (only its primary ranges)
Set<TokenRange> ranges2 = tokenMap.getTokenRanges(CqlIdentifier.fromCql("ks2"), node1);
// [Murmur3TokenRange(Murmur3Token(2), Murmur3Token(4)),
// Murmur3TokenRange(Murmur3Token(4), Murmur3Token(6)),
// Murmur3TokenRange(Murmur3Token(8), Murmur3Token(10)),
// Murmur3TokenRange(Murmur3Token(10), Murmur3Token(12))]
// (its primary ranges, and a replica of the primary ranges of node3, the previous node on the ring)
You can disable token metadata globally from the configuration:
datastax-java-driver.advanced.metadata.token-map.enabled = false
If it is disabled at startup, Metadata#getTokenMap will stay empty, and token-aware routing won’t work (requests will be sent to a non-optimal coordinator). If you disable it at runtime, it will keep the value of the last refresh, and token-aware routing might operate on stale data.
The keyspace-specific information in TokenMap (all methods with a CqlIdentifier argument) relies
on schema metadata. If schema metadata is disabled or filtered, token metadata will
also be unavailable for the excluded keyspaces.
Was this page helpful?
ScyllaDB Java Driver is available under the Apache v2 License. ScyllaDB Java Driver is a fork of DataStax Java Driver. See Copyright here.