pg: distributed named process groups

Replacement for pg2 module. Differences (compared to pg2):
 * non-existent and empty group treated the same (empty list of pids),
   thus create/1 and delete/1 have no effect (and not implemented).
   which_groups() return only non-empty groups
 * no cluster lock required, and no dependency on global
 * all join/leave operations require local process (it's not possible to join
   a process from a different node)
 * multi-join: join/leave several processes with a single call

Empty groups are not supported: unlike a process, group does not have
originating node. So it's possible that during net split one node deletes
the group, that still exists for another partition. pg2 will re-create deleted
group as soon as net split converges, which is quite unexpected.

Process groups can be organised into multiple scopes. Scopes are
completely independent of each other. A process may join any
number of groups in any number of scopes. Scopes are designed to
decouple single mesh into a set of overlay networks, reducing
amount of traffic required to propagate group membership
information.

1 files changed, 16 insertions, 3 deletions

diff --git a/lib/kernel/src/kernel.erl b/lib/kernel/src/kernel.erl
index 8877ceea8e..83f3fbecd5 100644
--- a/lib/kernel/src/kernel.erl
+++ b/lib/kernel/src/kernel.erl
@@ -64,7 +64,7 @@ config_change(Changed, New, Removed) ->
 %%%   (file,code,  | erl_dist (A)| | safe_sup (1)|
 %%%    rpc, ...)    -------------   -------------
 %%%		          |               |
-%%%                  (net_kernel,  (disk_log, pg2,
+%%%                  (net_kernel,  (disk_log, pg,
 %%%          	      auth, ...)     ...)
 %%%
 %%% The rectangular boxes are supervisors.  All supervisors except
@@ -180,7 +180,7 @@ init(safe) ->
 
     Boot = start_boot_server(),
     DiskLog = start_disk_log(),
-    Pg2 = start_pg2(),
+    Pg = start_pg2() ++ start_pg(),
 
     %% Run the on_load handlers for all modules that have been
     %% loaded so far. Running them at this point means that
@@ -188,7 +188,7 @@ init(safe) ->
     %% (and in particular call code:priv_dir/1 or code:lib_dir/1).
     init:run_on_load_handlers(),
 
-    {ok, {SupFlags, Boot ++ DiskLog ++ Pg2}}.
+    {ok, {SupFlags, Boot ++ DiskLog ++ Pg}}.
 
 start_distribution() ->
     Rpc = #{id => rex,
@@ -279,6 +279,19 @@ start_disk_log() ->
             []
     end.
 
+start_pg() ->
+    case application:get_env(kernel, start_pg) of
+        {ok, true} ->
+            [#{id => pg,
+                start => {pg, start_link, []},
+                restart => permanent,
+                shutdown => 1000,
+                type => worker,
+                modules => [pg]}];
+        _ ->
+            []
+    end.
+
 start_pg2() ->
     case application:get_env(kernel, start_pg2) of
         {ok, true} ->