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
|
// Test that removing and re-adding shard works correctly.
seedString = function(replTest) {
members = replTest.getReplSetConfig().members.map(function(elem) { return elem.host; });
return replTest.name + '/' + members.join(',');
};
removeShard = function(st, replTest) {
print( "Removing shard with name: " + replTest.name );
res = st.admin.runCommand( { removeshard: replTest.name } )
printjson(res);
assert( res.ok , "failed to start draining shard" );
checkRemoveShard = function() {
res = st.admin.runCommand( { removeshard: replTest.name } );
printjson(res);
return res.ok && res.msg == 'removeshard completed successfully';
}
assert.soon( checkRemoveShard , "failed to remove shard" );
// Need to wait for migration to be over... only works for inline deletes
checkNSLock = function() {
printjson( st.s.getDB( "config" ).locks.find().toArray() )
return !st.isAnyBalanceInFlight();
}
assert.soon( checkNSLock, "migrations did not end?" )
sleep( 2000 );
var directdb = replTest.getPrimary().getDB( "admin" );
assert.soon( function(){
var res = directdb.currentOp( { desc: /^clean/ } );
print( "eliot: " + replTest.getPrimary() + "\t" + tojson(res) );
return res.inprog.length == 0;
}, "never clean", 5 * 60 * 1000, 1000 );
replTest.getPrimary().getDB( coll.getDB().getName() ).dropDatabase();
print( "Shard removed successfully" );
};
addShard = function(st, replTest) {
seed = seedString(replTest);
print( "Adding shard with seed: " + seed );
try {
assert.eq(true, st.adminCommand({ addshard : seed }));
} catch (e) {
print("First attempt to addShard failed, trying again")
// transport error on first attempt is expected. Make sure second attempt goes through
assert.eq(true, st.adminCommand({ addshard : seed }));
}
ReplSetTest.awaitRSClientHosts( new Mongo( st.s.host ),
replTest.getSecondaries(),
{ok : true, secondary : true} );
assert.soon( function() {
var x = st.chunkDiff( coll.getName() , coll.getDB().getName() );
print( "chunk diff: " + x );
return x < 2;
} , "no balance happened", 5 * 60 * 1000 );
try {
assert.eq( 300, coll.find().itcount() );
} catch (e) {
// Expected. First query might get transport error and need to reconnect.
printjson(e);
assert.eq( 300, coll.find().itcount() );
}
print( "Shard added successfully" );
};
var st = new ShardingTest( testName = "remove2",
numShards = 2,
verboseLevel = 0,
numMongos = 1,
{ chunkSize : 1,
rs : true,
rs0 : { nodes : 2 },
rs1 : { nodes : 2 }
});
// Pending resolution of SERVER-8598, we need to wait for deletion after chunk migrations to avoid
// a pending delete re-creating a database after it was dropped.
st.s.getDB("config").settings.update( { _id: "balancer" },
{ $set : { _waitForDelete : true } },
true );
var rst0 = st._rs[0].test;
var rst1 = st._rs[1].test;
var conn = new Mongo( st.s.host );
var coll = conn.getCollection( "test.remove2" );
coll.drop();
// Decrease how long it will take for rst0 to time out its ReplicaSetMonitor for rst1 when rs1 is shut down
for( var i = 0; i < rst0.nodes.length; i++ ) {
node = rst0.nodes[i];
res = node.getDB('admin').runCommand({ setParameter : 1, replMonitorMaxFailedChecks : 1 });
printjson( res );
assert( res.ok );
}
st.admin.runCommand({ enableSharding : coll.getDB().getName() });
st.admin.runCommand({ shardCollection : coll.getFullName(), key: { i : 1 }});
// Setup initial data
var str = 'a';
while( str.length < 1024 * 16 ) {
str += str;
}
var bulk = coll.initializeUnorderedBulkOp();
for( var i = 0; i < 300; i++ ){
bulk.insert({ i: i % 10, str: str });
}
assert.writeOK(bulk.execute());
assert.eq( 300, coll.find().itcount() );
assert.soon( function() {
var x = st.chunkDiff( 'remove2' , "test" ); print( "chunk diff: " + x ); return x < 2;
} , "no balance happened", 5 * 60 * 1000 );
assert.eq( 300, coll.find().itcount() );
st.admin.printShardingStatus();
// Remove shard and add it back in, without shutting it down.
jsTestLog( "Attempting to remove shard and add it back in" )
removeShard( st, rst1 );
addShard(st, rst1 );
// Remove shard, restart set, then add it back in.
jsTestLog( "Attempting to remove shard, restart the set, and then add it back in" )
originalSeed = seedString(rst1);
removeShard( st, rst1 );
rst1.stopSet();
print( "Sleeping for 20 seconds to let the other shard's ReplicaSetMonitor time out" );
sleep( 20000 ); // 1 failed check should take 10 seconds, sleep for 20 just to be safe
rst1.startSet();
rst1.initiate();
rst1.awaitReplication();
assert.eq( originalSeed, seedString(rst1), "Set didn't come back up with the same hosts as before" );
addShard( st, rst1 );
// Shut down shard and wait for its ReplicaSetMonitor to be cleaned up, then start it back up and use it.
// TODO: test this both with AND without waiting for the ReplicaSetMonitor to be cleaned up.
// This part doesn't pass, even without cleaning up the ReplicaSetMonitor - see SERVER-5900.
/*printjson( conn.getDB('admin').runCommand({movePrimary : 'test2', to : rst1.name}) );
printjson( conn.getDB('admin').runCommand({setParameter : 1, replMonitorMaxFailedChecks : 5}) );
jsTestLog( "Shutting down set" )
rst1.stopSet();
jsTestLog( "sleeping for 20 seconds to make sure ReplicaSetMonitor gets cleaned up");
sleep(20000); // 1 failed check should take 10 seconds, sleep for 20 just to be safe
// Should fail since rst1 is the primary for test2
assert.throws(function() {conn.getDB('test2').foo.find().itcount()});
jsTestLog( "Bringing set back up" );
rst1.startSet();
rst1.initiate();
rst1.awaitReplication();
jsTestLog( "Checking that set is usable again" );
//conn.getDB('admin').runCommand({flushRouterConfig:1}); // Uncommenting this makes test pass
conn.getDB('test2').foo.insert({a:1});
gle = conn.getDB('test2').runCommand('getLastError');
if ( !gle.ok ) {
// Expected. First write will fail and need to re-connect
print( "write failed" );
printjson( gle );
conn.getDB('test2').foo.insert({a:1});
assert( conn.getDB('test2').getLastErrorObj().ok );
}
assert.eq( 1, conn.getDB('test2').foo.find().itcount() );
assert( conn.getDB('test2').dropDatabase().ok );*/
// Remove shard and add a new shard with the same replica set and shard name, but different ports.
jsTestLog( "Attempt removing shard and adding a new shard with the same Replica Set name" );
removeShard( st, rst1 );
rst1.stopSet();
print( "Sleeping for 20 seconds to let the other shard's ReplicaSetMonitor time out" );
sleep( 20000 );
var rst2 = new ReplSetTest({name : rst1.name, nodes : 2, startPort : rst1.startPort + 1500, useHostName : true});
rst2.startSet();
rst2.initiate();
rst2.awaitReplication();
addShard( st, rst2 );
printjson( st.admin.runCommand({movePrimary : 'test2', to : rst2.name}) );
assert.eq( 300, coll.find().itcount() );
conn.getDB('test2').foo.insert({a:1});
assert.eq( 1, conn.getDB('test2').foo.find().itcount() );
// Can't shut down with rst2 in the set or ShardingTest will fail trying to cleanup on shutdown.
// Have to take out rst2 and put rst1 back into the set so that it can clean up.
jsTestLog( "Putting ShardingTest back to state it expects" );
printjson( st.admin.runCommand({movePrimary : 'test2', to : rst0.name}) );
removeShard( st, rst2 );
rst2.stopSet();
rst1.startSet();
rst1.initiate();
rst1.awaitReplication();
assert.eq( originalSeed, seedString(rst1), "Set didn't come back up with the same hosts as before" );
addShard( st, rst1 );
jsTestLog( "finishing!" )
st.stop()
|
