From 34a186689a318ebc6fe2afaff396364d832bb21b Mon Sep 17 00:00:00 2001
From: Anders Svensson <anders@erlang.org>
Date: Sun, 6 Nov 2016 22:20:35 +0100
Subject: Don't assume nodes are eternally connected when sharing transport

Service configuration share_peers and use_shared_peers is used to share
peer connections with other connected nodes having a service of the same
name: a service process asks its neighbours about existing connections
when it starts, and pushes new connections as they're established.

The problem is that the mechanics assume that nodes() doesn't change. In
particular, if a neighbour isn't connected when a service starts then it
doesn't receive the request to share connections. Solve by having each
service process monitor nodes, a nodeup notification causing it to
request connections of its neighbours. Nodes going down is already
handled, by remote connections being monitored in diameter_service.
---
 lib/diameter/src/base/diameter_service.erl | 25 +++++++++++++++++++++++--
 1 file changed, 23 insertions(+), 2 deletions(-)

(limited to 'lib')

diff --git a/lib/diameter/src/base/diameter_service.erl b/lib/diameter/src/base/diameter_service.erl
index a976a8b998..788f697627 100644
--- a/lib/diameter/src/base/diameter_service.erl
+++ b/lib/diameter/src/base/diameter_service.erl
@@ -514,6 +514,13 @@ transition({tc_timeout, T}, S) ->
     tc_timeout(T, S),
     ok;
 
+transition({nodeup, Node, _}, S) ->
+    nodeup(Node, S),
+    ok;
+
+transition({nodedown, _Node, _}, _) ->
+    ok;
+
 transition(Req, S) ->
     unexpected(handle_info, [Req], S),
     ok.
@@ -709,6 +716,8 @@ mref(P) ->
 
 init_shared(#state{options = #{use_shared_peers := T},
                    service_name = Svc}) ->
+    T == false orelse net_kernel:monitor_nodes(true, [{node_type, visible},
+                                                      nodedown_reason]),
     notify(T, Svc, {service, self()}).
 
 init_mod(#diameter_app{alias = Alias,
@@ -728,6 +737,11 @@ notify(Share, SvcName, T) ->
 %% Test for the empty list for upgrade reasons: there's no
 %% diameter_peer:notify/3 in old code.
 
+nodeup(Node, #state{options = #{share_peers := SP},
+                    service_name = SvcName}) ->
+    lists:member(Node, remotes(SP))
+        andalso diameter_peer:notify([Node], SvcName, {service, self()}).
+
 remotes(false) ->
     [];
 
@@ -1400,9 +1414,15 @@ is_remote(Pid, T) ->
 %% # remote_peer_up/4
 %% ---------------------------------------------------------------------------
 
-remote_peer_up(TPid, Aliases, Caps, #state{options = #{use_shared_peers := T}}
+remote_peer_up(TPid, Aliases, Caps, #state{options = #{use_shared_peers := T},
+                                           remote = {PeerT, _, _}}
                                     = S) ->
-    is_remote(TPid, T) andalso rpu(TPid, Aliases, Caps, S).
+    is_remote(TPid, T)
+        andalso not ets:member(PeerT, TPid)
+        andalso rpu(TPid, Aliases, Caps, S).
+
+%% Notification can be duplicate since remote nodes push and the local
+%% node pulls.
 
 rpu(TPid, Aliases, Caps, #state{service = Svc, remote = RT}) ->
     #diameter_service{applications = Apps} = Svc,
@@ -1412,6 +1432,7 @@ rpu(TPid, Aliases, Caps, #state{service = Svc, remote = RT}) ->
 
 rpu(_, [] = No, _, _) ->
     No;
+
 rpu(TPid, Aliases, Caps, {PeerT, _, _} = RT) ->
     monitor(process, TPid),
     ets:insert(PeerT, #peer{pid = TPid,
-- 
cgit v1.2.3