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
|
/**
* This test ensures that a secondary that has gone "too stale" (i.e. cannot find another node with
* a common oplog point) will transition to RECOVERING state, stay in RECOVERING after restart, and
* transition back to SECONDARY once it finds a sync source with a common oplog point.
*
* Note: This test requires persistence in order for a restarted node with a stale oplog to stay in
* the RECOVERING state. A restarted node with an ephemeral storage engine will not have an oplog
* upon restart, so will immediately resync.
*
* @tags: [requires_persistence]
*
* Replica Set Setup:
*
* Node 0 (PRIMARY) : Small Oplog
* Node 1 (SECONDARY) : Large Oplog
* Node 2 (SECONDARY) : Small Oplog
*
* 1: Insert one document on the primary (Node 0) and ensure it is replicated.
* 2: Stop node 2.
* 3: Wait until Node 2 is down.
* 4: Overflow the primary's oplog.
* 5: Stop Node 1 and restart Node 2.
* 6: Wait for Node 2 to transition to RECOVERING (it should be too stale).
* 7: Stop and restart Node 2.
* 8: Wait for Node 2 to transition to RECOVERING (its oplog should remain stale after restart).
* 9: Restart Node 1, which should have the full oplog history.
* 10: Wait for Node 2 to leave RECOVERING and transition to SECONDARY.
*
*/
(function() {
"use strict";
load('jstests/replsets/rslib.js');
function getFirstOplogEntry(conn) {
return conn.getDB('local').oplog.rs.find().sort({$natural: 1}).limit(1)[0];
}
/**
* Overflows the oplog of a given node.
*
* To detect oplog overflow, we continuously insert large documents until we
* detect that the first entry of the oplog is no longer the same as when we started. This
* implies that the oplog attempted to grow beyond its maximum size i.e. it
* has overflowed/rolled over.
*
* Each document will be inserted with a writeConcern given by 'writeConcern'.
*
*/
function overflowOplog(conn, db, writeConcern) {
var firstOplogEntry = getFirstOplogEntry(primary);
var collName = "overflow";
// Keep inserting large documents until the oplog rolls over.
const largeStr = new Array(32 * 1024).join('aaaaaaaa');
while (bsonWoCompare(getFirstOplogEntry(conn), firstOplogEntry) === 0) {
assert.commandWorked(
db[collName].insert({data: largeStr}, {writeConcern: {w: writeConcern}}));
}
}
/**
* True if a node's entry in "members" has tooStale: true.
*/
function tooStale(conn) {
return assert.commandWorked(conn.adminCommand("replSetGetStatus")).tooStale;
}
/**
* Returns a node's current replica state.
*/
function myState(conn) {
return assert.commandWorked(conn.adminCommand("replSetGetStatus")).myState;
}
var testName = "too_stale_secondary";
var smallOplogSizeMB = 1;
var bigOplogSizeMB = 1000;
// Node 0 is given a small oplog so we can overflow it. Node 1's large oplog allows it to
// store all entries comfortably without overflowing, so that Node 2 can eventually use it as
// a sync source after it goes too stale. Because this test overflows the oplog, a small
// syncdelay is chosen to frequently take checkpoints, allowing oplog truncation to proceed.
var replTest = new ReplSetTest({
name: testName,
nodes:
[{oplogSize: smallOplogSizeMB}, {oplogSize: bigOplogSizeMB}, {oplogSize: smallOplogSizeMB}],
nodeOptions: {syncdelay: 1},
});
var nodes = replTest.startSet();
replTest.initiate({
_id: testName,
members: [
{_id: 0, host: nodes[0].host},
{_id: 1, host: nodes[1].host, priority: 0},
{_id: 2, host: nodes[2].host, priority: 0}
]
});
var dbName = testName;
var collName = "test";
jsTestLog("Wait for Node 0 to become the primary.");
replTest.waitForState(replTest.nodes[0], ReplSetTest.State.PRIMARY);
var primary = replTest.getPrimary();
var primaryTestDB = primary.getDB(dbName);
jsTestLog("1: Insert one document on the primary (Node 0) and ensure it is replicated.");
assert.commandWorked(primaryTestDB[collName].insert({a: 1}, {writeConcern: {w: 3}}));
assert(!tooStale(replTest.nodes[2]));
jsTestLog("2: Stop Node 2.");
replTest.stop(2);
jsTestLog("3: Wait until Node 2 is down.");
replTest.waitForState(replTest.nodes[2], ReplSetTest.State.DOWN);
var firstOplogEntryNode1 = getFirstOplogEntry(replTest.nodes[1]);
jsTestLog("4: Overflow the primary's oplog.");
overflowOplog(primary, primaryTestDB, 2);
// Make sure that Node 1's oplog didn't overflow.
assert.eq(firstOplogEntryNode1,
getFirstOplogEntry(replTest.nodes[1]),
"Node 1's oplog overflowed unexpectedly.");
jsTestLog("5: Stop Node 1 and restart Node 2.");
replTest.stop(1);
replTest.restart(2);
jsTestLog("6: Wait for Node 2 to transition to RECOVERING (it should be too stale).");
assert.soonNoExcept(() => tooStale(replTest.nodes[2]), "Waiting for Node 2 to become too stale");
// This checks the state as reported by the node itself.
assert.soon(() => myState(replTest.nodes[2]) === ReplSetTest.State.RECOVERING,
"Waiting for Node 2 to transition to RECOVERING");
// This waits for the state as indicated by the primary node.
replTest.waitForState(replTest.nodes[2], ReplSetTest.State.RECOVERING);
jsTestLog("7: Stop and restart Node 2.");
replTest.stop(2);
replTest.restart(2);
jsTestLog(
"8: Wait for Node 2 to transition to RECOVERING (its oplog should remain stale after restart)");
assert.soonNoExcept(() => tooStale(replTest.nodes[2]), "Waiting for Node 2 to become too stale");
// This checks the state as reported by the node itself.
assert.soon(() => myState(replTest.nodes[2]) === ReplSetTest.State.RECOVERING,
"Waiting for Node 2 to transition to RECOVERING");
// This waits for the state as indicated by the primary node.
replTest.waitForState(replTest.nodes[2], ReplSetTest.State.RECOVERING);
jsTestLog("9: Restart Node 1, which should have the full oplog history.");
replTest.restart(1);
jsTestLog("10: Wait for Node 2 to leave RECOVERING and transition to SECONDARY.");
assert.soonNoExcept(() => !tooStale(replTest.nodes[2]), "Waiting for Node 2 to exit too stale");
// This checks the state as reported by the node itself.
assert.soon(() => myState(replTest.nodes[2]) === ReplSetTest.State.SECONDARY,
"Waiting for Node 2 to transition to SECONDARY");
// This waits for the state as indicated by the primary node.
replTest.waitForState(replTest.nodes[2], ReplSetTest.State.SECONDARY);
replTest.stopSet();
}());
|
