base/frameworks/cluster/pools.zeek

Cluster

Defines an interface for managing pools of cluster nodes. Pools are a useful way to distribute work or data among nodes within a cluster.

Namespace

Cluster

Imports

base/frameworks/cluster/main.zeek, base/utils/hash_hrw.zeek

Summary

State Variables

Cluster::logger_pool: Cluster::Pool

A pool containing all the logger nodes of a cluster.

Cluster::logger_pool_spec: Cluster::PoolSpec &redef

The specification for Cluster::logger_pool.

Cluster::proxy_pool: Cluster::Pool

A pool containing all the proxy nodes of a cluster.

Cluster::proxy_pool_spec: Cluster::PoolSpec &redef

The specification for Cluster::proxy_pool.

Cluster::worker_pool: Cluster::Pool

A pool containing all the worker nodes of a cluster.

Cluster::worker_pool_spec: Cluster::PoolSpec &redef

The specification for Cluster::worker_pool.

Types

Cluster::PoolNode: record

Store state of a cluster within the context of a work pool.

Cluster::PoolNodeTable: table

Cluster::PoolSpec: record

A pool specification.

Cluster::RoundRobinTable: table

Functions

Cluster::hrw_topic: function

Retrieve the topic associated with the node mapped via Rendezvous hash of an arbitrary key.

Cluster::register_pool: function

Registers and initializes a pool.

Cluster::rr_log_topic: function

Distributes log message topics among logger nodes via round-robin.

Cluster::rr_topic: function

Retrieve the topic associated with the node in a round-robin fashion.

Detailed Interface

State Variables

Cluster::logger_pool
Type

Cluster::Pool

Default
{
   spec=[topic=<uninitialized>, node_type=<uninitialized>, max_nodes=<uninitialized>, exclusive=F]
   nodes={

   }
   node_list=[]
   hrw_pool=[sites={

   }]
   rr_key_seq={

   }
   alive_count=0
}

A pool containing all the logger nodes of a cluster. The pool’s node membership/availability is automatically maintained by the cluster framework.

Cluster::logger_pool_spec
Type

Cluster::PoolSpec

Attributes

&redef

Default
{
   topic="zeek/cluster/pool/logger"
   node_type=Cluster::LOGGER
   max_nodes=<uninitialized>
   exclusive=F
}

The specification for Cluster::logger_pool.

Cluster::proxy_pool
Type

Cluster::Pool

Default
{
   spec=[topic=<uninitialized>, node_type=<uninitialized>, max_nodes=<uninitialized>, exclusive=F]
   nodes={

   }
   node_list=[]
   hrw_pool=[sites={

   }]
   rr_key_seq={

   }
   alive_count=0
}

A pool containing all the proxy nodes of a cluster. The pool’s node membership/availability is automatically maintained by the cluster framework.

Cluster::proxy_pool_spec
Type

Cluster::PoolSpec

Attributes

&redef

Default
{
   topic="zeek/cluster/pool/proxy"
   node_type=Cluster::PROXY
   max_nodes=<uninitialized>
   exclusive=F
}

The specification for Cluster::proxy_pool.

Cluster::worker_pool
Type

Cluster::Pool

Default
{
   spec=[topic=<uninitialized>, node_type=<uninitialized>, max_nodes=<uninitialized>, exclusive=F]
   nodes={

   }
   node_list=[]
   hrw_pool=[sites={

   }]
   rr_key_seq={

   }
   alive_count=0
}

A pool containing all the worker nodes of a cluster. The pool’s node membership/availability is automatically maintained by the cluster framework.

Cluster::worker_pool_spec
Type

Cluster::PoolSpec

Attributes

&redef

Default
{
   topic="zeek/cluster/pool/worker"
   node_type=Cluster::WORKER
   max_nodes=<uninitialized>
   exclusive=F
}

The specification for Cluster::worker_pool.

Types

Cluster::PoolNode
Type

record

name: string

The node name (e.g. “manager”).

alias: string

An alias of name used to prevent hashing collisions when creating site_id.

site_id: count

A 32-bit unique identifier for the pool node, derived from name/alias.

alive: bool &default = F &optional

Whether the node is currently alive and can receive work.

Store state of a cluster within the context of a work pool.

Cluster::PoolNodeTable
Type

table [string] of Cluster::PoolNode

Cluster::PoolSpec
Type

record

topic: string

A topic string that can be used to reach all nodes within a pool.

node_type: Cluster::NodeType

The type of nodes that are contained within the pool.

max_nodes: count &optional

The maximum number of nodes that may belong to the pool. If not set, then all available nodes will be added to the pool, else the cluster framework will automatically limit the pool membership according to the threshold.

exclusive: bool &default = F &optional

Whether the pool requires exclusive access to nodes. If true, then max_nodes nodes will not be assigned to any other pool. When using this flag, max_nodes must also be set.

A pool specification.

Cluster::RoundRobinTable
Type

table [string] of int

Functions

Cluster::hrw_topic
Type

function (pool: Cluster::Pool, key: any) : string

Retrieve the topic associated with the node mapped via Rendezvous hash of an arbitrary key.

Parameters
  • pool – the pool of nodes to consider.

  • key – data used for input to the hashing function that will uniformly distribute keys among available nodes.

Returns

a topic string associated with a cluster node that is alive or an empty string if nothing is alive.

Cluster::register_pool
Type

function (spec: Cluster::PoolSpec) : Cluster::Pool

Registers and initializes a pool.

Cluster::rr_log_topic
Type

function (id: Log::ID, path: string) : string

Distributes log message topics among logger nodes via round-robin. This will be automatically assigned to Broker::log_topic if Cluster::enable_round_robin_logging is enabled. If no logger nodes are active, then this will return the value of Broker::default_log_topic.

Cluster::rr_topic
Type

function (pool: Cluster::Pool, key: string &default = "" &optional) : string

Retrieve the topic associated with the node in a round-robin fashion.

Parameters
  • pool – the pool of nodes to consider.

  • key – an arbitrary string to identify the purpose for which you’re requesting the topic. e.g. consider using a name-spaced key like “Intel::cluster_rr_key” if you need to guarantee that a group of messages get distributed in a well-defined pattern without other messages being interleaved within the round-robin. Usually sharing the default key is fine for load-balancing purposes.

Returns

a topic string associated with a cluster node that is alive, or an empty string if nothing is alive.