Add docs for ClusterManager and WorkerConfig

stdlib/Distributed/docs/src/index.md

@@ -60,6 +60,8 @@ same host, and `SSHManager`, for launching on remote hosts via `ssh`. TCP/IP soc
 and transport messages between processes. It is possible for Cluster Managers to provide a different transport.
 
 ```@docs
+Distributed.ClusterManager
+Distributed.WorkerConfig
 Distributed.launch
 Distributed.manage
 Distributed.kill(::ClusterManager, ::Int, ::WorkerConfig)

stdlib/Distributed/src/cluster.jl

@@ -1,7 +1,46 @@
 # This file is a part of Julia. License is MIT: https://julialang.org/license
 
+"""
+    ClusterManager
+
+Supertype for cluster managers, which control workers processes as a cluster.
+Cluster managers implement how workers can be added, removed and communicated with.
+`SSHManager` and `LocalManager` are subtypes of this.
+"""
 abstract type ClusterManager end
 
+"""
+    WorkerConfig
+
+Type used by [`ClusterManager`](@ref)s to control workers added to their clusters. Some fields
+are used by all cluster managers to access a host:
+  * `io` -- the connection used to access the worker (a subtype of `IO` or `Nothing`)
+  * `host` -- the host address (either an `AbstractString` or `Nothing`)
+  * `port` -- the port on the host used to connect to the worker (either an `Int` or `Nothing`)
+
+Some are used by the cluster manager to add workers to an already-initialized host:
+  * `count` -- the number of workers to be launched on the host
+  * `exename` -- the path to the Julia executable on the host, defaults to `"\$(Sys.BINDIR)/julia"` or
+    `"\$(Sys.BINDIR)/julia-debug"`
+  * `exeflags` -- flags to use when lauching Julia remotely
+
+The `userdata` field is used to store information for each worker by external managers.
+
+Some fields are used by `SSHManager` and similar managers:
+  * `tunnel` -- `true` (use tunneling), `false` (do not use tunneling), or [`nothing`](@ref) (use default for the manager)
+  * `bind_addr` -- the address on the remote host to bind to
+  * `sshflags` -- flags to use in establishing the SSH connection
+  * `max_parallel` -- the maximum number of workers to connect to in parallel on the host
+
+Some fields are used by both `LocalManager`s and `SSHManager`s:
+  * `connect_at` -- determines whether this is a worker-to-worker or driver-to-worker setup call
+  * `process` -- the process which will be connected (usually the manager will assign this during [`addprocs`](@ref))
+  * `ospid` -- the process ID according to the host OS, used to interrupt worker processes
+  * `environ` -- private dictionary used to store temporary information by Local/SSH managers
+  * `ident` -- worker as identified by the [`ClusterManager`](@ref)
+  * `connect_idents` -- list of worker ids the worker must connect to if using a custom topology
+  * `enable_threaded_blas` -- `true`, `false`, or `nothing`, whether to use threaded BLAS or not on the workers
+"""
 mutable struct WorkerConfig
     # Common fields relevant to all cluster managers
     io::Union{IO, Nothing}