Skip to content

Commit c765132

Browse files
author
Emile Joubert
committed
Merged bug22877 into default
2 parents 279c8ed + 6a1a950 commit c765132

File tree

4 files changed

+117
-53
lines changed

4 files changed

+117
-53
lines changed

‎docs/rabbitmqctl.1.xml‎

Lines changed: 27 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -270,8 +270,8 @@
270270
<title>Cluster management</title>
271271

272272
<variablelist>
273-
<varlistentry>
274-
<term><cmdsynopsis><command>cluster</command> <arg choice="req" role="usage-option-list"><replaceable>clusternode</replaceable> ...</arg></cmdsynopsis></term>
273+
<varlistentry id="cluster">
274+
<term><cmdsynopsis><command>cluster</command><arg choice="req" role="usage-option-list"><replaceable>clusternode</replaceable> ...</arg></cmdsynopsis></term>
275275
<listitem>
276276
<variablelist>
277277
<varlistentry>
@@ -281,7 +281,8 @@
281281
</variablelist>
282282
<para>
283283
Instruct the node to become member of a cluster with the
284-
specified nodes.
284+
specified nodes. To cluster with currently offline nodes,
285+
use <link linkend="force_cluster"><command>force_cluster</command></link>.
285286
</para>
286287
<para>
287288
Cluster nodes can be of two types: disk or ram. Disk nodes
@@ -334,6 +335,29 @@
334335
</para>
335336
</listitem>
336337
</varlistentry>
338+
<varlistentry id="force_cluster">
339+
<term><cmdsynopsis><command>force_cluster</command><arg choice="req" role="usage-option-list"><replaceable>clusternode</replaceable> ...</arg></cmdsynopsis></term>
340+
<listitem>
341+
<variablelist>
342+
<varlistentry>
343+
<term>clusternode</term>
344+
<listitem><para>Subset of the nodes of the cluster to which this node should be connected.</para></listitem>
345+
</varlistentry>
346+
</variablelist>
347+
<para>
348+
Instruct the node to become member of a cluster with the
349+
specified nodes. This will succeed even if the specified nodes
350+
are offline. For a more detailed description, see
351+
<link linkend="cluster"><command>cluster</command>.</link>
352+
</para>
353+
<para>
354+
Note that this variant of the cluster command just
355+
ignores the current status of the specified nodes.
356+
Clustering may still fail for a variety of other
357+
reasons.
358+
</para>
359+
</listitem>
360+
</varlistentry>
337361
</variablelist>
338362
</refsect2>
339363

‎src/rabbit_control.erl‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -160,6 +160,12 @@ action(cluster, Node, ClusterNodeSs, Inform) ->
160160
[Node, ClusterNodes]),
161161
rpc_call(Node, rabbit_mnesia, cluster, [ClusterNodes]);
162162

163+
action(force_cluster, Node, ClusterNodeSs, Inform) ->
164+
ClusterNodes = lists:map(fun list_to_atom/1, ClusterNodeSs),
165+
Inform("Forcefully clustering node ~p with ~p (ignoring offline nodes)",
166+
[Node, ClusterNodes]),
167+
rpc_call(Node, rabbit_mnesia, force_cluster, [ClusterNodes]);
168+
163169
action(status, Node, [], Inform) ->
164170
Inform("Status of node ~p", [Node]),
165171
case call(Node, {rabbit, status, []}) of

‎src/rabbit_mnesia.erl‎

Lines changed: 62 additions & 36 deletions
Original file line numberDiff line numberDiff line change
@@ -33,8 +33,8 @@
3333
-module(rabbit_mnesia).
3434

3535
-export([ensure_mnesia_dir/0, dir/0, status/0, init/0, is_db_empty/0,
36-
cluster/1, reset/0, force_reset/0, is_clustered/0,
37-
empty_ram_only_tables/0]).
36+
cluster/1, force_cluster/1, reset/0, force_reset/0,
37+
is_clustered/0, empty_ram_only_tables/0]).
3838

3939
-export([table_names/0]).
4040

@@ -58,6 +58,8 @@
5858
-spec(init/0 :: () -> 'ok').
5959
-spec(is_db_empty/0 :: () -> boolean()).
6060
-spec(cluster/1 :: ([node()]) -> 'ok').
61+
-spec(force_cluster/1 :: ([node()]) -> 'ok').
62+
-spec(cluster/2 :: ([node()], boolean()) -> 'ok').
6163
-spec(reset/0 :: () -> 'ok').
6264
-spec(force_reset/0 :: () -> 'ok').
6365
-spec(is_clustered/0 :: () -> boolean()).
@@ -88,24 +90,30 @@ status() ->
8890
init() ->
8991
ok = ensure_mnesia_running(),
9092
ok = ensure_mnesia_dir(),
91-
ok = init_db(read_cluster_nodes_config()),
93+
ok = init_db(read_cluster_nodes_config(), true),
9294
ok = wait_for_tables(),
9395
ok.
9496

9597
is_db_empty() ->
9698
lists:all(fun (Tab) -> mnesia:dirty_first(Tab) == '$end_of_table' end,
9799
table_names()).
98100

101+
cluster(ClusterNodes) ->
102+
cluster(ClusterNodes, false).
103+
force_cluster(ClusterNodes) ->
104+
cluster(ClusterNodes, true).
105+
99106
%% Alter which disk nodes this node is clustered with. This can be a
100107
%% subset of all the disk nodes in the cluster but can (and should)
101108
%% include the node itself if it is to be a disk rather than a ram
102-
%% node.
103-
cluster(ClusterNodes) ->
109+
%% node. If Force is false, only connections to online nodes are
110+
%% allowed.
111+
cluster(ClusterNodes, Force) ->
104112
ok = ensure_mnesia_not_running(),
105113
ok = ensure_mnesia_dir(),
106114
rabbit_misc:ensure_ok(mnesia:start(), cannot_start_mnesia),
107115
try
108-
ok = init_db(ClusterNodes),
116+
ok = init_db(ClusterNodes, Force),
109117
ok = wait_for_tables(),
110118
ok = create_cluster_nodes_config(ClusterNodes)
111119
after
@@ -277,38 +285,56 @@ delete_cluster_nodes_config() ->
277285

278286
%% Take a cluster node config and create the right kind of node - a
279287
%% standalone disk node, or disk or ram node connected to the
280-
%% specified cluster nodes.
281-
init_db(ClusterNodes) ->
282-
case mnesia:change_config(extra_db_nodes, ClusterNodes -- [node()]) of
283-
{ok, []} ->
284-
case mnesia:system_info(use_dir) of
285-
true ->
286-
case check_schema_integrity() of
287-
ok ->
288-
ok;
289-
{error, Reason} ->
290-
%% NB: we cannot use rabbit_log here since
291-
%% it may not have been started yet
292-
error_logger:warning_msg(
293-
"schema integrity check failed: ~p~n"
294-
"moving database to backup location "
295-
"and recreating schema from scratch~n",
296-
[Reason]),
297-
ok = move_db(),
288+
%% specified cluster nodes. If Force is false, don't allow
289+
%% connections to offline nodes.
290+
init_db(ClusterNodes, Force) ->
291+
UClusterNodes = lists:usort(ClusterNodes),
292+
ProperClusterNodes = UClusterNodes -- [node()],
293+
case mnesia:change_config(extra_db_nodes, ProperClusterNodes) of
294+
{ok, Nodes} ->
295+
case Force of
296+
false ->
297+
FailedClusterNodes = ProperClusterNodes -- Nodes,
298+
case FailedClusterNodes of
299+
[] -> ok;
300+
_ ->
301+
throw({error, {failed_to_cluster_with,
302+
FailedClusterNodes,
303+
"Mnesia could not connect to some nodes."}})
304+
end;
305+
_ -> ok
306+
end,
307+
case Nodes of
308+
[] ->
309+
case mnesia:system_info(use_dir) of
310+
true ->
311+
case check_schema_integrity() of
312+
ok ->
313+
ok;
314+
{error, Reason} ->
315+
%% NB: we cannot use rabbit_log here since
316+
%% it may not have been started yet
317+
error_logger:warning_msg(
318+
"schema integrity check failed: ~p~n"
319+
"moving database to backup location "
320+
"and recreating schema from scratch~n",
321+
[Reason]),
322+
ok = move_db(),
323+
ok = create_schema()
324+
end;
325+
false ->
298326
ok = create_schema()
299327
end;
300-
false ->
301-
ok = create_schema()
302-
end;
303-
{ok, [_|_]} ->
304-
IsDiskNode = ClusterNodes == [] orelse
305-
lists:member(node(), ClusterNodes),
306-
ok = wait_for_replicated_tables(),
307-
ok = create_local_table_copy(schema, disc_copies),
308-
ok = create_local_table_copies(case IsDiskNode of
309-
true -> disc;
310-
false -> ram
311-
end);
328+
[_|_] ->
329+
IsDiskNode = ClusterNodes == [] orelse
330+
lists:member(node(), ClusterNodes),
331+
ok = wait_for_replicated_tables(),
332+
ok = create_local_table_copy(schema, disc_copies),
333+
ok = create_local_table_copies(case IsDiskNode of
334+
true -> disc;
335+
false -> ram
336+
end)
337+
end;
312338
{error, Reason} ->
313339
%% one reason we may end up here is if we try to join
314340
%% nodes together that are currently running standalone or

‎src/rabbit_tests.erl‎

Lines changed: 22 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -597,27 +597,27 @@ test_cluster_management() ->
597597

598598
ok = control_action(reset, []),
599599
lists:foreach(fun (Arg) ->
600-
ok = control_action(cluster, Arg),
600+
ok = control_action(force_cluster, Arg),
601601
ok
602602
end,
603603
ClusteringSequence),
604604
lists:foreach(fun (Arg) ->
605605
ok = control_action(reset, []),
606-
ok = control_action(cluster, Arg),
606+
ok = control_action(force_cluster, Arg),
607607
ok
608608
end,
609609
ClusteringSequence),
610610
ok = control_action(reset, []),
611611
lists:foreach(fun (Arg) ->
612-
ok = control_action(cluster, Arg),
612+
ok = control_action(force_cluster, Arg),
613613
ok = control_action(start_app, []),
614614
ok = control_action(stop_app, []),
615615
ok
616616
end,
617617
ClusteringSequence),
618618
lists:foreach(fun (Arg) ->
619619
ok = control_action(reset, []),
620-
ok = control_action(cluster, Arg),
620+
ok = control_action(force_cluster, Arg),
621621
ok = control_action(start_app, []),
622622
ok = control_action(stop_app, []),
623623
ok
@@ -628,13 +628,13 @@ test_cluster_management() ->
628628
ok = control_action(reset, []),
629629
ok = control_action(start_app, []),
630630
ok = control_action(stop_app, []),
631-
ok = control_action(cluster, ["invalid1@invalid",
632-
"invalid2@invalid"]),
631+
ok = control_action(force_cluster, ["invalid1@invalid",
632+
"invalid2@invalid"]),
633633

634634
%% join a non-existing cluster as a ram node
635635
ok = control_action(reset, []),
636-
ok = control_action(cluster, ["invalid1@invalid",
637-
"invalid2@invalid"]),
636+
ok = control_action(force_cluster, ["invalid1@invalid",
637+
"invalid2@invalid"]),
638638

639639
SecondaryNode = rabbit_misc:makenode("hare"),
640640
case net_adm:ping(SecondaryNode) of
@@ -659,27 +659,35 @@ test_cluster_management2(SecondaryNode) ->
659659

660660
%% join cluster as a ram node
661661
ok = control_action(reset, []),
662-
ok = control_action(cluster, [SecondaryNodeS, "invalid1@invalid"]),
662+
ok = control_action(force_cluster, [SecondaryNodeS, "invalid1@invalid"]),
663663
ok = control_action(start_app, []),
664664
ok = control_action(stop_app, []),
665665

666666
%% change cluster config while remaining in same cluster
667-
ok = control_action(cluster, ["invalid2@invalid", SecondaryNodeS]),
667+
ok = control_action(force_cluster, ["invalid2@invalid", SecondaryNodeS]),
668668
ok = control_action(start_app, []),
669669
ok = control_action(stop_app, []),
670670

671671
%% join non-existing cluster as a ram node
672-
ok = control_action(cluster, ["invalid1@invalid",
673-
"invalid2@invalid"]),
672+
ok = control_action(force_cluster, ["invalid1@invalid",
673+
"invalid2@invalid"]),
674+
ok = control_action(start_app, []),
675+
ok = control_action(stop_app, []),
676+
677+
%% join empty cluster as a ram node
678+
ok = control_action(cluster, []),
679+
ok = control_action(start_app, []),
680+
ok = control_action(stop_app, []),
681+
674682
%% turn ram node into disk node
675683
ok = control_action(reset, []),
676684
ok = control_action(cluster, [SecondaryNodeS, NodeS]),
677685
ok = control_action(start_app, []),
678686
ok = control_action(stop_app, []),
679687

680688
%% convert a disk node into a ram node
681-
ok = control_action(cluster, ["invalid1@invalid",
682-
"invalid2@invalid"]),
689+
ok = control_action(force_cluster, ["invalid1@invalid",
690+
"invalid2@invalid"]),
683691

684692
%% turn a disk node into a ram node
685693
ok = control_action(reset, []),

0 commit comments

Comments
 (0)