[snmp] Tweaking agent start sequence to reduce error reporting

<This is a hack>
Tweaked the agent start sequence in order to minimize error
reporting during a failed start (due to net-if issues).
</This is a hack>
Also added a new net-if option in order to change what
otherwise will be an error reporet to an info report.

OTP-18422 (ERIERL-873)

6 files changed, 146 insertions, 51 deletions

diff --git a/lib/snmp/src/agent/snmpa_agent.erl b/lib/snmp/src/agent/snmpa_agent.erl
index 5039b08391..b70bf310d3 100644
--- a/lib/snmp/src/agent/snmpa_agent.erl
+++ b/lib/snmp/src/agent/snmpa_agent.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %%
-%% Copyright Ericsson AB 1996-2020. All Rights Reserved.
+%% Copyright Ericsson AB 1996-2023. All Rights Reserved.
 %%
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -349,10 +349,16 @@ init([Prio, Parent, Ref, Options]) ->
        "~n   Options: ~p", [Prio, Parent, Ref, Options]),
     case (catch do_init(Prio, Parent, Ref, Options)) of
 	{ok, State} ->
-	    ?vdebug("started",[]),
+	    ?vdebug("started"),
 	    {ok, State};
+	{error, {net_if, info, Reason}} ->
+	    info_msg("Failed starting agent: "
+                     "~n   Net If error: ~p", [Reason]),
+            %% {shutdown, Reason};
+            exit(Reason);
 	{error, Reason} ->
-	    config_err("failed starting agent: ~n~p", [Reason]),
+	    config_err("Failed starting agent: "
+                       "~n   ~p", [Reason]),
 	    {stop, Reason}
     end.
 
@@ -419,15 +425,18 @@ start_note_store(Prio, Ref, Options) ->
 	{ok, Pid} ->
 	    ?vdebug("start_note_store -> Pid: ~p", [Pid]),
 	    Pid;
-	{error, Reason} -> 
-	    ?vinfo("error starting note store: ~n~p",[Reason]),
-	    throw({error, {note_store_error, Reason}});
-	{'EXIT', Reason} ->
-	    ?vinfo("exit starting note store: ~n~p",[Reason]),
-	    throw({error, {note_store_exit, Reason}});
+	{error, {Reason, _ChildSpec}} -> 
+	    ?vinfo("error starting note store: "
+                   "~n   ~p", [Reason]),
+	    throw({error, {note_store, error, Reason}});
+	{'EXIT', {Reason, _ChildSpec}} ->
+	    ?vinfo("exit starting note store: "
+                   "~n   ~p", [Reason]),
+	    throw({error, {note_store, exit, Reason}});
 	Error ->
-	    ?vinfo("failed starting note store: ~n~p",[Error]),
-	    throw({error, {note_store_failed, Error}})
+	    ?vinfo("failed starting note store: "
+                   "~n   ~p", [Error]),
+	    throw({error, {note_store, failed, Error}})
     end.
     
 
@@ -446,18 +455,25 @@ start_net_if(none, Prio, Ref, Vsns, NoteStore, Options) ->
 
     case (catch snmpa_misc_sup:start_net_if(Prio, NoteStore, Ref, self(),
 					    Mod, NiOpts)) of
-	{ok, Pid} -> 
+	{ok, Pid} ->
 	    ?vdebug("start_net_if -> Pid: ~p", [Pid]),
 	    {master_agent, Pid, Mod};
-	{error, Reason} -> 
-	    ?vinfo("error starting net if: ~n~p",[Reason]),
-	    throw({error, {net_if_error, Reason}});
+	{error, {{Class, udp_open, PortNo, Reason}, _ChildSpec}} ->
+	    ?vinfo("error starting net if: "
+                   "~n   ~p", [Reason]),
+	    throw({error, {net_if, Class, {udp_open, PortNo, Reason}}});
+	{error, {Reason, _ChildSpec}} ->
+	    ?vinfo("error starting net if: "
+                   "~n   ~p", [Reason]),
+	    throw({error, {net_if, error, Reason}});
 	{'EXIT', Reason} ->
-	    ?vinfo("exit starting net if: ~n~p",[Reason]),
-	    throw({error, {net_if_exit, Reason}});
+	    ?vinfo("exit starting net if: "
+                   "~n   ~p", [Reason]),
+	    throw({error, {net_if, exit, Reason}});
 	Error ->
-	    ?vinfo("failed starting net if: ~n~p",[Error]),
-	    throw({error, {net_if_failed, Error}})
+	    ?vinfo("failed starting net if: "
+                   "~n   ~p", [Error]),
+	    throw({error, {net_if, failed, Error}})
     end;
 start_net_if(Parent, _Prio, _Ref, _Vsns, _NoteStore, _Options) 
   when is_pid(Parent) ->
@@ -479,15 +495,18 @@ start_mib_server(Prio, Ref, Mibs, Options) ->
 	{ok, Pid} ->
 	    ?vdebug("start_mib_server -> Pid: ~p", [Pid]),
 	    Pid;
-	{error, Reason} -> 
-	    ?vinfo("error starting mib server: ~n~p",[Reason]),
-	    throw({error, {mib_server_error, Reason}});
+	{error, {Reason, _ChildSpec}} ->
+	    ?vinfo("error starting mib server: "
+                   "~n   ~p", [Reason]),
+	    throw({error, {mib_server, error, Reason}});
 	{'EXIT', Reason} ->
-	    ?vinfo("exit starting mib server: ~n~p",[Reason]),
-	    throw({error, {mib_server_exit, Reason}});
+	    ?vinfo("exit starting mib server: "
+                   "~n   ~p", [Reason]),
+	    throw({error, {mib_server, exit, Reason}});
 	Error ->
-	    ?vinfo("failed starting mib server: ~n~p",[Error]),
-	    throw({error, {mib_server_failed, Error}})
+	    ?vinfo("failed starting mib server: "
+                   "~n   ~p", [Error]),
+	    throw({error, {mib_server, failed, Error}})
     end.
 
 
@@ -3462,8 +3481,8 @@ get_stats_counters([Counter|Counters], Acc) ->
 
 %% ---------------------------------------------------------------------
 
-%% info_msg(F, A) ->
-%%     ?snmpa_info(F, A).
+info_msg(F, A) ->
+    ?snmpa_info(F, A).
 
 warning_msg(F, A) ->
     ?snmpa_warning(F, A).
diff --git a/lib/snmp/src/agent/snmpa_agent_sup.erl b/lib/snmp/src/agent/snmpa_agent_sup.erl
index 0a5116b2d0..b1eeaba988 100644
--- a/lib/snmp/src/agent/snmpa_agent_sup.erl
+++ b/lib/snmp/src/agent/snmpa_agent_sup.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %% 
-%% Copyright Ericsson AB 1997-2016. All Rights Reserved.
+%% Copyright Ericsson AB 1997-2023. All Rights Reserved.
 %% 
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -24,7 +24,9 @@
 -behaviour(supervisor).
 
 %% External exports
--export([start_link/0, start_link/1, start_subagent/3, stop_subagent/1]).
+-export([start_link/0, start_link/1,
+         start_master_agent/1,
+         start_subagent/3, stop_subagent/1]).
 
 %% Internal exports
 -export([init/1]).
@@ -40,8 +42,7 @@
 
 
 %%%-----------------------------------------------------------------
-%%% This is a supervisor for the mib processes.  Each agent has one
-%%% mib process.
+%%% This is a supervisor for the agent processes (master and sub).
 %%%-----------------------------------------------------------------
 start_link() ->
     ?d("start_link -> entry", []),
@@ -52,6 +53,9 @@ start_link(AgentSpec) ->
 	"~n   AgentSpec: ~p", [AgentSpec]),
     supervisor:start_link({local, ?SERVER}, ?MODULE, [[AgentSpec]]).
 
+start_master_agent(MasterAgentSpec) ->
+    supervisor:start_child(snmpa_agent_sup, MasterAgentSpec).
+
 start_subagent(ParentAgent, Subtree, Mibs) ->
     ?d("start_subagent -> entry with"
 	"~n   ParentAgent: ~p"
diff --git a/lib/snmp/src/agent/snmpa_net_if.erl b/lib/snmp/src/agent/snmpa_net_if.erl
index bd19a32159..2c726eedad 100644
--- a/lib/snmp/src/agent/snmpa_net_if.erl
+++ b/lib/snmp/src/agent/snmpa_net_if.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %% 
-%% Copyright Ericsson AB 2004-2022. All Rights Reserved.
+%% Copyright Ericsson AB 2004-2023. All Rights Reserved.
 %% 
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -235,11 +235,21 @@ init(Prio, NoteStore, MasterAgent, Parent, Opts) ->
 		    end,
 		    erlang:raise(C, E, S)
 	    end;
-	{error, Reason} ->
-	    config_err("failed starting net-if: ~n~p", [Reason]),
-	    proc_lib:init_ack({error, Reason});
+	{error, {udp_open, {open, PortNo, Reason}}} ->
+            OEFilters = get_open_err_filters(Opts),
+            Class =
+                case lists:member(Reason, OEFilters) of
+                    false ->
+                        error;
+                    true ->
+                        info
+            end,
+            proc_lib:init_ack({error, {Class, udp_open, PortNo, Reason}});
+        {error, Reason} ->
+            %% config_err("failed starting net-if: ~n~p", [Reason]),
+            proc_lib:init_ack({error, Reason});
 	Error ->
-	    config_err("failed starting net-if: ~n~p", [Error]),
+	    %% config_err("failed starting net-if: ~n~p", [Error]),
 	    proc_lib:init_ack({error, Error})
     end.
 
@@ -256,7 +266,7 @@ do_init(Prio, NoteStore, MasterAgent, Parent, Opts) ->
     Vsns = get_vsns(Opts),
     ?vdebug("vsns: ~w",[Vsns]),
 
-    %% Flow control --
+    %% -- Flow control --
     Limit      = get_req_limit(Opts),
     ?vdebug("Limit: ~w", [Limit]),
     FilterOpts = get_filter_opts(Opts),
@@ -475,7 +485,7 @@ gen_udp_open(system, Opts) ->
                     throw({udp_open, {port, PReason}})
             end;
 	{error, OReason} ->
-            throw({udp_open, {open, OReason}})
+            throw({udp_open, {open, 0, OReason}})
     end;
 %% This is for "future compat" since we cannot actually config '0'...
 gen_udp_open(IpPort, Opts) when (IpPort =:= 0) ->
@@ -533,7 +543,7 @@ gen_udp_range_open(Min, Max, Opts) ->
             gen_udp_range_open(Min+1, Max, Opts);
         {error, Reason} ->
             ?vdebug("gen_udp_range_open(~w,~w) -> ~w", [Reason]),
-            throw({udp_open, {open, Reason}})
+            throw({udp_open, {open, Min, Reason}})
     catch
         C:E:S ->
             ?vinfo("gen_udp_range_open(~w,~w) -> failed open socket: "
@@ -2104,6 +2114,16 @@ get_filter_opts(O) ->
 get_filter_module(O) ->
     snmp_misc:get_option(module, O, ?DEFAULT_FILTER_MODULE).
 
+get_open_err_filters(O) ->
+    case snmp_misc:get_option(open_err_filters, O, []) of
+        Filters when is_list(Filters) ->
+            Filters;
+        Filter when is_atom(Filter) ->
+            [Filter];
+        _ ->
+            []
+    end.
+
 get_recbuf(Opts, DefaultOpts) -> 
     get_socket_opt(recbuf, Opts, DefaultOpts, use_default).
 
@@ -2156,8 +2176,8 @@ info_msg(F,A) ->
 user_err(F, A) ->
     snmpa_error:user_err(F, A).
  
-config_err(F, A) ->
-    snmpa_error:config_err(F, A).
+%% config_err(F, A) ->
+%%     snmpa_error:config_err(F, A).
  
 
 %% ----------------------------------------------------------------
diff --git a/lib/snmp/src/agent/snmpa_supervisor.erl b/lib/snmp/src/agent/snmpa_supervisor.erl
index 7d5c6da2c8..3261411ec3 100644
--- a/lib/snmp/src/agent/snmpa_supervisor.erl
+++ b/lib/snmp/src/agent/snmpa_supervisor.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %% 
-%% Copyright Ericsson AB 1996-2019. All Rights Reserved.
+%% Copyright Ericsson AB 1996-2023. All Rights Reserved.
 %% 
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -29,6 +29,7 @@
 %% Internal exports
 -export([init/1, config/2]).
 
+-compile({no_auto_import,[erase/1]}).
 
 -define(SERVER, ?MODULE).
 
@@ -41,6 +42,10 @@
 %% Process structure
 %% =================
 %%
+%%                                "application"
+%%                                     |
+%%                                  app-sup
+%%                                     |
 %%             ___________________ supervisor __________________
 %%            /              |              |          \         \ 
 %%   ___misc_sup___    target_cache  symbolic_store   local_db   agent_sup
@@ -176,7 +181,23 @@ start_master_sup(Opts) ->
 
 do_start_master_sup(Opts) ->
     verify_mandatory([db_dir], Opts),
-    supervisor:start_link({local, ?SERVER}, ?MODULE, [master, Opts]).  
+    case supervisor:start_link({local, ?SERVER}, ?MODULE, [master, Opts]) of
+        {ok, Pid} = OK ->
+            %% <HACKETI-HACK-HACK>
+            Key = master_agent_child_spec,
+            MasterAgentSpec = lookup(Key),
+            case snmpa_agent_sup:start_master_agent(MasterAgentSpec) of
+                {ok, MPid} when is_pid(MPid) ->
+                    erase(Key),
+                    OK;
+                {error, {Reason, _ChildSpec}} ->
+                    stop(Pid, 0),
+                    {error, Reason}
+            end;
+            %% </HACKETI-HACK-HACK>
+        Else ->
+            Else
+    end.
 
 verify_mandatory([], _) ->
     ok;
@@ -503,9 +524,18 @@ init([AgentType, Opts]) ->
 		    worker_spec(snmpa_agent, 
 				[Prio, snmp_master_agent, none, Ref, AgentOpts],
 				Restart, 15000),
+                %% <HACKETI-HACK-HACK>
+                %% The point is to make start failure more quiet
+                %% Often the failure happens in the master agent,
+                %% so we move the start of that out of this function
+                %% and into the 'do_start_master_sup' function.
+                %% At some point we should rewrite this. Maybe start all
+                %% children the same way (explicitly).
+                store(master_agent_child_spec, AgentSpec),
 		AgentSupSpec = 
-		    sup_spec(snmpa_agent_sup, [AgentSpec], 
+		    sup_spec(snmpa_agent_sup, [], 
 			     Restart, infinity), 
+                %% </HACKETI-HACK-HACK>
 		[ConfigSpec, AgentSupSpec];
 	    _ ->
 		?vdebug("[sub agent] spec for the agent supervisor",[]),
@@ -521,6 +551,13 @@ init([AgentType, Opts]) ->
 store(Key, Value) ->
     ets:insert(snmp_agent_table, {Key, Value}).
 
+lookup(Key) ->
+    [{Key, Value}] = ets:lookup(snmp_agent_table, Key),
+    Value.
+
+erase(Key) ->
+    ets:delete(snmp_agent_table, Key).
+
 get_mibs(Mibs, Vsns) ->
     MibDir = filename:join(code:priv_dir(snmp), "mibs"),
     StdMib = 
diff --git a/lib/snmp/src/app/snmp_app.erl b/lib/snmp/src/app/snmp_app.erl
index 486b276383..7ba3c659f0 100644
--- a/lib/snmp/src/app/snmp_app.erl
+++ b/lib/snmp/src/app/snmp_app.erl
@@ -1,7 +1,7 @@
 %% 
 %% %CopyrightBegin%
 %% 
-%% Copyright Ericsson AB 2003-2021. All Rights Reserved.
+%% Copyright Ericsson AB 2003-2023. All Rights Reserved.
 %% 
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -38,8 +38,13 @@ start(Type, []) ->
     %% First start the (new) central supervisor,
     {ok, Pid} = snmp_app_sup:start_link(),
     Entities = entities(),
-    ok = start_entities(Type, Entities),
-    {ok, Pid}.
+    case start_entities(Type, Entities) of
+        ok ->
+            {ok, Pid};
+        Error ->
+            snmp_app_sup:stop(),
+            Error
+    end.
 
 entities() ->
     entities([agent, manager], []).
diff --git a/lib/snmp/src/app/snmp_app_sup.erl b/lib/snmp/src/app/snmp_app_sup.erl
index eb89cc5b6d..bb6faa18cb 100644
--- a/lib/snmp/src/app/snmp_app_sup.erl
+++ b/lib/snmp/src/app/snmp_app_sup.erl
@@ -1,7 +1,7 @@
 %% 
 %% %CopyrightBegin%
 %% 
-%% Copyright Ericsson AB 2003-2016. All Rights Reserved.
+%% Copyright Ericsson AB 2003-2023. All Rights Reserved.
 %% 
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -64,7 +64,12 @@ start_agent(Type, Opts) ->
 	"~n   Type: ~p"
 	"~n   Opts: ~p", [Type, Opts]),
     Restart = get_restart(Opts, permanent), 
-    start_sup_child(snmpa_supervisor, Restart, [Type, Opts]).
+    case start_sup_child(snmpa_supervisor, Restart, [Type, Opts]) of
+        {ok, Pid} = OK when is_pid(Pid) ->
+            OK;
+        {error, {Reason, _ChildSpec}} ->
+            {error, Reason}
+    end.
 
 
 start_manager(Type, Opts) ->
@@ -72,7 +77,12 @@ start_manager(Type, Opts) ->
 	"~n   Type: ~p"
 	"~n   Opts: ~p", [Type, Opts]),
     Restart = get_restart(Opts, transient), 
-    start_sup_child(snmpm_supervisor, Restart, [Type, Opts]).
+    case start_sup_child(snmpm_supervisor, Restart, [Type, Opts]) of
+        {ok, Pid} = OK when is_pid(Pid) ->
+            OK;
+        {error, {Reason, _ChildSpec}} ->
+            {error, Reason}
+    end.
 
 
 %%%-------------------------------------------------------------------