module Mongo::Cluster::Topology

Defines behavior for getting servers.

Topologies are associated with their clusters - for example, a ReplicaSet topology contains the replica set name. A topology object therefore cannot be used with multiple cluster objects.

At the same time, topology objects do not know anything about specific servers in a cluster, despite what their constructor may suggest. Which means, in particular, that topology change events require the application to maintain cluster references on its own if it wants to track server changes within a replica set.

@since 2.0.0

Constants

OPTIONS

The various topologies for server selection.

@since 2.0.0 @api private

Public Instance Methods

initial(cluster, monitoring, options) click to toggle source

Get the initial cluster topology for the provided options.

@example Get the initial cluster topology.

Topology.initial(topology: :replica_set)

@param [ Cluster ] cluster The cluster. @param [ Monitoring ] monitoring The monitoring. @param [ Hash ] options The cluster options.

@option options [ true | false ] :direct_connection Whether to connect

directly to the specified seed, bypassing topology discovery. Exactly
one seed must be provided.

@option options [ Symbol ] :connect Deprecated - use :direct_connection

option instead of this option. The connection method to use. This
forces the cluster to behave in the specified way instead of
auto-discovering. One of :direct, :replica_set, :sharded,
:load_balanced. If :connect is set to :load_balanced, the driver
will behave as if the server is a load balancer even if it isn't
connected to a load balancer.

@option options [ true | false ] :load_balanced Whether to expect to

connect to a load balancer.

@option options [ Symbol ] :replica_set The name of the replica set to

connect to. Servers not in this replica set will be ignored.

@return [ ReplicaSet, Sharded, Single, LoadBalanced ] The topology.

@since 2.0.0 @api private

# File lib/mongo/cluster/topology.rb, line 90
def initial(cluster, monitoring, options)
  connect = options[:connect]&.to_sym
  cls = if options[:direct_connection]
    if connect && connect != :direct
      raise ArgumentError, "Conflicting topology options: direct_connection=true and connect=#{connect}"
    end
    if options[:load_balanced]
      raise ArgumentError, "Conflicting topology options: direct_connection=true and load_balanced=true"
    end
    Single
  elsif options[:direct_connection] == false && connect && connect == :direct
    raise ArgumentError, "Conflicting topology options: direct_connection=false and connect=#{connect}"
  elsif connect && connect != :load_balanced
    if options[:load_balanced]
      raise ArgumentError, "Conflicting topology options: connect=#{options[:connect].inspect} and load_balanced=true"
    end
    OPTIONS.fetch(options[:connect].to_sym)
  elsif options.key?(:replica_set) || options.key?(:replica_set_name)
    if options[:load_balanced]
      raise ArgumentError, "Conflicting topology options: replica_set/replica_set_name and load_balanced=true"
    end
    ReplicaSetNoPrimary
  elsif options[:load_balanced] || connect == :load_balanced
    LoadBalanced
  else
    Unknown
  end
  # Options here are client/cluster/server options.
  # In particular the replica set name key is different for
  # topology.
  # If replica_set_name is given (as might be internally by driver),
  # use that key.
  # Otherwise (e.g. options passed down from client),
  # move replica_set to replica_set_name.
  if (cls <= ReplicaSetNoPrimary || cls == Single) && !options[:replica_set_name]
    options = options.dup
    options[:replica_set_name] = options.delete(:replica_set)
  end
  cls.new(options, monitoring, cluster)
end