Clarify what single shard means and that leader shards are placed on one node (#361)

Follower shards are also placed on one node

Author: Simran-B

site/content/3.10/deploy/oneshard.md

@@ -10,14 +10,20 @@ archetype: default
 {{< tag "ArangoDB Enterprise Edition" "ArangoGraph" >}}
 
 The OneShard option for ArangoDB clusters restricts all collections of a
-database to a single shard and places them on one DB-Server node. This way,
-whole queries can be pushed to and executed on that server, massively reducing
-cluster-internal communication. The Coordinator only gets back the final result.
+database to a single shard so that every collection has `numberOfShards` set to `1`,
+and all leader shards are placed on one DB-Server node. This way, whole queries
+can be pushed to and executed on that server, massively reducing cluster-internal
+communication. The Coordinator only gets back the final result.
 
 Queries are always limited to a single database, and with the data of a whole
 database on a single node, the OneShard option allows running transactions with
 ACID guarantees on shard leaders.
 
+Collections can have replicas by setting a `replicationFactor` greater than `1`
+as usual. For each replica, the follower shards are all placed on one DB-Server
+node when using the OneShard option. This allows for a quick failover in case
+the DB-Server with the leader shards fails.
+
 A OneShard setup is highly recommended for most graph use cases and join-heavy
 queries.
 

site/content/3.11/deploy/oneshard.md

@@ -10,14 +10,20 @@ archetype: default
 {{< tag "ArangoDB Enterprise Edition" "ArangoGraph" >}}
 
 The OneShard option for ArangoDB clusters restricts all collections of a
-database to a single shard and places them on one DB-Server node. This way,
-whole queries can be pushed to and executed on that server, massively reducing
-cluster-internal communication. The Coordinator only gets back the final result.
+database to a single shard so that every collection has `numberOfShards` set to `1`,
+and all leader shards are placed on one DB-Server node. This way, whole queries
+can be pushed to and executed on that server, massively reducing cluster-internal
+communication. The Coordinator only gets back the final result.
 
 Queries are always limited to a single database, and with the data of a whole
 database on a single node, the OneShard option allows running transactions with
 ACID guarantees on shard leaders.
 
+Collections can have replicas by setting a `replicationFactor` greater than `1`
+as usual. For each replica, the follower shards are all placed on one DB-Server
+node when using the OneShard option. This allows for a quick failover in case
+the DB-Server with the leader shards fails.
+
 A OneShard setup is highly recommended for most graph use cases and join-heavy
 queries.
 

site/content/3.12/deploy/oneshard.md

@@ -10,14 +10,20 @@ archetype: default
 {{< tag "ArangoDB Enterprise Edition" "ArangoGraph" >}}
 
 The OneShard option for ArangoDB clusters restricts all collections of a
-database to a single shard and places them on one DB-Server node. This way,
-whole queries can be pushed to and executed on that server, massively reducing
-cluster-internal communication. The Coordinator only gets back the final result.
+database to a single shard so that every collection has `numberOfShards` set to `1`,
+and all leader shards are placed on one DB-Server node. This way, whole queries
+can be pushed to and executed on that server, massively reducing cluster-internal
+communication. The Coordinator only gets back the final result.
 
 Queries are always limited to a single database, and with the data of a whole
 database on a single node, the OneShard option allows running transactions with
 ACID guarantees on shard leaders.
 
+Collections can have replicas by setting a `replicationFactor` greater than `1`
+as usual. For each replica, the follower shards are all placed on one DB-Server
+node when using the OneShard option. This allows for a quick failover in case
+the DB-Server with the leader shards fails.
+
 A OneShard setup is highly recommended for most graph use cases and join-heavy
 queries.