1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
|
/**
* Check that on a loss of primary, another node doesn't assume primary if it is stale. We force a
* stepDown to test this.
*
* This test also checks that the serverStatus command metrics replSetStepDown and
* replSetStepDownWithForce are incremented correctly.
*
* This test requires the fsync command to force a secondary to be stale.
* @tags: [requires_fsync]
*/
load("jstests/replsets/libs/election_metrics.js");
load("jstests/replsets/rslib.js");
// We are bypassing collection validation because this test runs "shutdown" command so the server is
// expected to be down when MongoRunner.stopMongod is called.
TestData.skipCollectionAndIndexValidation = true;
var replTest = new ReplSetTest({
name: 'testSet',
nodes: {"n0": {rsConfig: {priority: 2}}, "n1": {}, "n2": {rsConfig: {votes: 1, priority: 0}}},
nodeOptions: {verbose: 1}
});
var nodes = replTest.startSet();
replTest.initiate();
replTest.waitForState(nodes[0], ReplSetTest.State.PRIMARY);
var primary = replTest.getPrimary();
// do a write
print("\ndo a write");
assert.commandWorked(primary.getDB("foo").bar.insert({x: 1}));
replTest.awaitReplication();
// In the event of any error, we have to unlock any nodes that we have fsyncLocked.
function unlockNodes(nodes) {
jsTestLog('Unlocking nodes: ' + tojson(nodes));
nodes.forEach(function(node) {
try {
jsTestLog('Unlocking node: ' + node);
assert.commandWorked(node.getDB("admin").fsyncUnlock());
} catch (e) {
jsTestLog('Failed to unlock node: ' + node + ': ' + tojson(e) +
'. Ignoring unlock error and moving on to next node.');
}
});
}
var lockedNodes = [];
try {
// lock secondaries
jsTestLog('Locking nodes: ' + tojson(replTest.getSecondaries()));
replTest.getSecondaries().forEach(function(node) {
jsTestLog('Locking node: ' + node);
jsTestLog(
'fsync lock ' + node + ' result: ' +
tojson(assert.commandWorked(node.getDB("admin").runCommand({fsync: 1, lock: 1}))));
lockedNodes.push(node);
});
jsTestLog('Stepping down primary: ' + primary);
for (var i = 0; i < 11; i++) {
// do another write
assert.commandWorked(primary.getDB("foo").bar.insert({x: i}));
}
let res = assert.commandWorked(primary.adminCommand({replSetGetStatus: 1}));
assert(res.electionCandidateMetrics,
() => "Response should have an 'electionCandidateMetrics' field: " + tojson(res));
let intitialServerStatus = assert.commandWorked(primary.adminCommand({serverStatus: 1}));
jsTestLog('Do stepdown of primary ' + primary + ' that should not work');
// this should fail, so we don't need to try/catch
jsTestLog(
'Step down ' + primary + ' expected error: ' +
tojson(assert.commandFailed(primary.getDB("admin").runCommand({replSetStepDown: 10}))));
// Check that the 'total' and 'failed' fields of 'replSetStepDown' have been incremented in
// serverStatus and that they have not been incremented for 'replSetStepDownWithForce'.
let newServerStatus = assert.commandWorked(primary.adminCommand({serverStatus: 1}));
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDown,
newServerStatus.metrics.commands.replSetStepDown,
"total",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDown,
newServerStatus.metrics.commands.replSetStepDown,
"failed",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDownWithForce,
newServerStatus.metrics.commands.replSetStepDownWithForce,
"total",
0);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDownWithForce,
newServerStatus.metrics.commands.replSetStepDownWithForce,
"failed",
0);
// This section checks that the metrics are incremented accurately when the command fails due to
// an error occurring before stepDown is called in the replication coordinator, such as due to
// bad values or type mismatches in the arguments, or checkReplEnabledForCommand returning a bad
// status. The stepdown period being negative is one example of such an error, but success in
// this case gives us confidence that the behavior in the other cases is the same.
// Stepdown should fail because the stepdown period is negative
jsTestLog('Do stepdown of primary ' + primary + ' that should not work');
assert.commandFailedWithCode(
primary.getDB("admin").runCommand({replSetStepDown: -1, force: true}), ErrorCodes.BadValue);
// Check that the 'total' and 'failed' fields of 'replSetStepDown' and
// 'replSetStepDownWithForce' have been incremented in serverStatus.
intitialServerStatus = newServerStatus;
newServerStatus = assert.commandWorked(primary.adminCommand({serverStatus: 1}));
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDown,
newServerStatus.metrics.commands.replSetStepDown,
"total",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDown,
newServerStatus.metrics.commands.replSetStepDown,
"failed",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDownWithForce,
newServerStatus.metrics.commands.replSetStepDownWithForce,
"total",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDownWithForce,
newServerStatus.metrics.commands.replSetStepDownWithForce,
"failed",
1);
jsTestLog('Do stepdown of primary ' + primary + ' that should work');
assert.commandWorked(
primary.adminCommand({replSetStepDown: ReplSetTest.kDefaultTimeoutMS, force: true}));
// Check that the 'total' fields of 'replSetStepDown' and 'replSetStepDownWithForce' have been
// incremented in serverStatus and that their 'failed' fields have not been incremented.
intitialServerStatus = newServerStatus;
newServerStatus = assert.commandWorked(primary.adminCommand({serverStatus: 1}));
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDown,
newServerStatus.metrics.commands.replSetStepDown,
"total",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDown,
newServerStatus.metrics.commands.replSetStepDown,
"failed",
0);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDownWithForce,
newServerStatus.metrics.commands.replSetStepDownWithForce,
"total",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDownWithForce,
newServerStatus.metrics.commands.replSetStepDownWithForce,
"failed",
0);
jsTestLog('Checking hello on ' + primary);
var r2 = assert.commandWorked(primary.getDB("admin").runCommand({hello: 1}));
jsTestLog('Result from running hello on ' + primary + ': ' + tojson(r2));
assert.eq(r2.isWritablePrimary, false);
assert.eq(r2.secondary, true);
// Check that the 'electionCandidateMetrics' section of the replSetGetStatus response has been
// cleared, since the node is no longer primary.
res = assert.commandWorked(primary.adminCommand({replSetGetStatus: 1}));
assert(!res.electionCandidateMetrics,
() => "Response should not have an 'electionCandidateMetrics' field: " + tojson(res));
// This section checks that the metrics are incremented accurately when the command fails due to
// an error while stepping down. This is one reason the replSetStepDown command could fail once
// we call stepDown in the replication coordinator, but success in this case gives us confidence
// that the behavior in the other cases is the same.
// Stepdown should fail because the node is no longer primary
jsTestLog('Do stepdown of primary ' + primary + ' that should not work');
assert.commandFailedWithCode(primary.getDB("admin").runCommand(
{replSetStepDown: ReplSetTest.kDefaultTimeoutMS, force: true}),
ErrorCodes.NotWritablePrimary);
// Check that the 'total' and 'failed' fields of 'replSetStepDown' and
// 'replSetStepDownWithForce' have been incremented in serverStatus.
intitialServerStatus = newServerStatus;
newServerStatus = assert.commandWorked(primary.adminCommand({serverStatus: 1}));
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDown,
newServerStatus.metrics.commands.replSetStepDown,
"total",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDown,
newServerStatus.metrics.commands.replSetStepDown,
"failed",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDownWithForce,
newServerStatus.metrics.commands.replSetStepDownWithForce,
"total",
1);
verifyServerStatusChange(intitialServerStatus.metrics.commands.replSetStepDownWithForce,
newServerStatus.metrics.commands.replSetStepDownWithForce,
"failed",
1);
} catch (e) {
throw e;
} finally {
unlockNodes(lockedNodes);
}
print("\nreset stepped down time");
assert.commandWorked(primary.getDB("admin").runCommand({replSetFreeze: 0}));
primary = replTest.getPrimary();
print("\nawait");
replTest.awaitSecondaryNodes(90000);
replTest.awaitReplication();
// 'n0' may have just voted for 'n1', preventing it from becoming primary for the first 30 seconds
// of this assert.soon
assert.soon(function() {
try {
var result = primary.getDB("admin").runCommand({hello: 1});
return new RegExp(":" + replTest.nodes[0].port + "$").test(result.primary);
} catch (x) {
return false;
}
}, "wait for n0 to be primary", 60000);
primary = replTest.getPrimary();
var firstPrimary = primary;
print("\nprimary is now " + firstPrimary);
assert.adminCommandWorkedAllowingNetworkError(primary, {replSetStepDown: 100, force: true});
print("\nget a primary");
replTest.getPrimary();
assert.soon(function() {
var secondPrimary = replTest.getPrimary();
return firstPrimary.host !== secondPrimary.host;
}, "making sure " + firstPrimary.host + " isn't still primary", 60000);
// Add arbiter for shutdown tests
replTest.add();
print("\ncheck shutdown command");
primary = replTest.getPrimary();
var secondary = replTest.getSecondary();
try {
secondary.adminCommand({shutdown: 1});
} catch (e) {
print(e);
}
primary = replTest.getPrimary();
assert.soon(function() {
try {
var result = primary.getDB("admin").runCommand({replSetGetStatus: 1});
for (var i in result.members) {
if (result.members[i].self) {
continue;
}
return result.members[i].health == 0;
}
} catch (e) {
print("error getting status from primary: " + e);
primary = replTest.getPrimary();
return false;
}
}, 'make sure primary knows that secondary is down before proceeding');
print("\nrunning shutdown without force on primary: " + primary);
// this should fail because the primary can't reach an up-to-date secondary (because the only
// secondary is down)
var now = new Date();
assert.commandFailed(primary.getDB("admin").runCommand({shutdown: 1, timeoutSecs: 3}));
// on windows, javascript and the server perceive time differently, to compensate here we use 2750ms
assert.gte((new Date()) - now, 2750);
print("\nsend shutdown command");
var currentPrimary = replTest.getPrimary();
try {
printjson(currentPrimary.getDB("admin").runCommand({shutdown: 1, force: true}));
} catch (e) {
if (!isNetworkError(e)) {
throw e;
}
}
print("checking " + currentPrimary + " is actually shutting down");
assert.soon(function() {
try {
currentPrimary.findOne();
} catch (e) {
return true;
}
return false;
});
print("\nOK 1");
replTest.stopSet();
print("OK 2");
|