summaryrefslogtreecommitdiff
path: root/jstests/sharding/chunk_history_window.js
blob: dc9d0b29cf84c38a2f7912f05968c306a2d6c297 (plain)
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
/**
 * Test that config servers keep chunk history for up to minSnapshotHistoryWindowInSeconds.
 *
 * @tags: [
 *   requires_fcv_47,
 *   requires_majority_read_concern,
 *   requires_persistence,
 * ]
 *
 * - Create a one-chunk sharded collection, its history is [{validAfter: T0}].
 * - Insert a document at timestamp insertTS.
 * - Move the chunk, its history is [{validAfter: T1}, {validAfter: T0}], where T1 > insertTS > T0.
 * - Until now > insertTS + window - margin, read at insertTS and assert success.
 * - After now > T0 + window + margin, T0 is expired. Move the chunk, triggering a history cleanup.
 * - History is [{validAfter: T2}, {validAfter: T1}], where T2 > T1 > insertTS > T0.
 * - Read at insertTS and assert failure with StaleChunkHistory.
 * - Read at T2 - 1 sec, assert success.
 */
(function() {
"use strict";

load("jstests/sharding/libs/sharded_transactions_helpers.js");

// The snapshot window is the max of minSnapshotHistoryWindowInSeconds and
// transactionLifetimeLimitSeconds.
const transactionLifetimeLimitSecs = 15;
const minSnapshotHistoryWindowSecs = transactionLifetimeLimitSecs;
const snapshotHistoryWindowSecs =
    Math.max(minSnapshotHistoryWindowSecs, transactionLifetimeLimitSecs);

const st = new ShardingTest({
    shards: {rs0: {nodes: 2}, rs1: {nodes: 2}},
    other: {
        configOptions: {
            setParameter: {
                minSnapshotHistoryWindowInSeconds: minSnapshotHistoryWindowSecs,
                transactionLifetimeLimitSeconds: transactionLifetimeLimitSecs,
                logComponentVerbosity: tojson({sharding: {verbosity: 2}})
            }
        },
        rsOptions: {
            setParameter: {
                minSnapshotHistoryWindowInSeconds: minSnapshotHistoryWindowSecs,
                transactionLifetimeLimitSeconds: transactionLifetimeLimitSecs,
            }
        }
    }
});

const primaryAdmin = st.rs0.getPrimary().getDB("admin");
assert.eq(assert
              .commandWorked(
                  primaryAdmin.runCommand({getParameter: 1, minSnapshotHistoryWindowInSeconds: 1}))
              .minSnapshotHistoryWindowInSeconds,
          minSnapshotHistoryWindowSecs);

const configAdmin = st.configRS.getPrimary().getDB("admin");
assert.eq(assert
              .commandWorked(
                  configAdmin.runCommand({getParameter: 1, minSnapshotHistoryWindowInSeconds: 1}))
              .minSnapshotHistoryWindowInSeconds,
          minSnapshotHistoryWindowSecs);

const mongosDB = st.s.getDB(jsTestName());
const mongosColl = mongosDB.test;
const ns = `${jsTestName()}.test`;

assert.commandWorked(mongosDB.adminCommand({enableSharding: mongosDB.getName()}));
st.ensurePrimaryShard(mongosDB.getName(), st.rs0.getURL());
st.shardColl(mongosColl, {_id: 1}, false);

const getChunkHistory = (query) => {
    // Always read from the config server primary in case there is a failover.
    const configChunks = st.configRS.getPrimary().getDB("config")["chunks"];
    return configChunks.findOne(query);
};

const origChunk = (function() {
    const coll = st.configRS.getPrimary().getDB("config").collections.findOne({_id: ns});
    if (coll.timestamp) {
        return getChunkHistory({uuid: coll.uuid});
    } else {
        return getChunkHistory({ns: ns});
    }
}());
jsTestLog(`Original chunk: ${tojson(origChunk)}`);
assert.eq(1, origChunk.history.length, tojson(origChunk));
let result = mongosDB.runCommand({insert: "test", documents: [{_id: 0}]});
const insertTS = assert.commandWorked(result).operationTime;
jsTestLog(`Inserted one document at ${tojson(insertTS)}`);
assert.lte(origChunk.history[0].validAfter, insertTS, `history: ${tojson(origChunk.history)}`);

jsTestLog("Move chunk to shard 1, create second history entry");
assert.commandWorked(st.s.adminCommand({moveChunk: ns, find: {_id: 0}, to: st.shard1.shardName}));
const postMoveChunkTime = Date.now();
let chunk = getChunkHistory({_id: origChunk._id});
jsTestLog(`Chunk: ${tojson(chunk)}`);
assert.eq(2, chunk.history.length, tojson(chunk));

// Test history window with 1s margin.
const testMarginMS = 1000;

// Test that reading from a snapshot at insertTS is valid for up to snapshotHistoryWindowSecs
// minus the testMarginMS (as a buffer).
const testWindowMS = snapshotHistoryWindowSecs * 1000 - testMarginMS;
while (Date.now() - 1000 * insertTS.getTime() < testWindowMS) {
    // Test that reading from a snapshot at insertTS is still valid.
    assert.commandWorked(mongosDB.runCommand(
        {find: "test", readConcern: {level: "snapshot", atClusterTime: insertTS}}));

    chunk = getChunkHistory({_id: origChunk._id});
    assert.eq(2, chunk.history.length, tojson(chunk));
    sleep(50);
}

// Sleep until our most recent chunk move is before the oldest history in our window.
const chunkExpirationTime = postMoveChunkTime + snapshotHistoryWindowSecs * 1000;
sleep(chunkExpirationTime + testMarginMS - Date.now());

jsTestLog("Move chunk back to shard 0 to trigger history cleanup");
assert.commandWorked(st.s.adminCommand({moveChunk: ns, find: {_id: 0}, to: st.shard0.shardName}));
chunk = getChunkHistory({_id: origChunk._id});
jsTestLog(`Chunk: ${tojson(chunk)}`);
// Oldest history entry was deleted: we added one and deleted one, still have two.
assert.eq(2, chunk.history.length, tojson(chunk));
assert.gte(chunk.history[1].validAfter, insertTS, `history: ${tojson(chunk.history)}`);

flushRoutersAndRefreshShardMetadata(st, {ns});

// Test that reading from a snapshot at insertTS returns StaleChunkHistory: the shards have enough
// history but the config servers don't.
assert.commandFailedWithCode(
    mongosDB.runCommand({find: "test", readConcern: {level: "snapshot", atClusterTime: insertTS}}),
    ErrorCodes.StaleChunkHistory);

// One second before the newest history entry is valid (check we don't delete *all* old entries).
var recentTS = Timestamp(chunk.history[0].validAfter.getTime() - 1, 0);
assert.commandWorked(
    mongosDB.runCommand({find: "test", readConcern: {level: "snapshot", atClusterTime: recentTS}}));

st.stop();
})();