Merge pull request #22 from davidoliverSP2/4.4-cluster-support-cp

Add section on cluster routing capability

Author: davidoliverSP2

modules/ROOT/pages/actions/index.adoc

@@ -10,6 +10,7 @@ There are several _actions_ that can be performed using the Cypher transaction H
 * xref:actions/transaction-flow.adoc[]
 * xref:actions/query-format.adoc[]
 * xref:actions/result-format.adoc[]
+* xref:actions/routing.adoc[]
 
 *Using the API:*
 
@@ -31,4 +32,3 @@ There are several _actions_ that can be performed using the Cypher transaction H
 * xref:actions/expired-transactions.adoc[]
 * xref:actions/handling-errors.adoc[]
 * xref:actions/handling-errors-in-an-open-transaction.adoc[]
-

modules/ROOT/pages/actions/routing.adoc

@@ -0,0 +1,25 @@
+:description: HTTP API routing.
+
+[[http-api-actions-cluster-routing]]
+= Clusters and routing
+
+The Cypher Transactional API provides a limited level of support for query routing.
+
+The level of support depends on whether link:https://neo4j.com/docs/operations-manual/current/clustering/internals/#clustering-routing[Server-side routing] has been enabled:
+
+[options="header", cols="20m,80a"]
+|===
+| Server-side routing
+| Support
+
+|disabled
+|Clusters support read-only queries.
+
+|enabled
+|Clusters support routing for queries that involve a xref::/actions/begin-and-commit-a-transaction-in-one-request.adoc[single request] (i.e. via the `tx/commit` endpoint).
+This is because the cluster does not currently support transaction identifiers across the cluster.
+
+In this context, all queries are considered `WRITE` queries, even if they contain no cypher with write operations.
+Since the HTTP API did not previously support marking queries as `READ` or `WRITE`, the server assumes all queries are `WRITE` queries to support all cases.
+The effect of this is that all queries are ultimately executed on the leader.
+|===