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
|
// Tests the reset logic for the periodic query yield timer. Regression test for SERVER-21341.
(function() {
'use strict';
var dbpath = MongoRunner.dataPath + jsTest.name();
resetDbpath(dbpath);
var mongod = MongoRunner.runMongod({dbpath: dbpath});
var coll = mongod.getDB("test").getCollection(jsTest.name());
// Configure the server so that queries are expected to yield after every 10 work cycles, or
// after every 500 milliseconds (whichever comes first). In addition, enable a failpoint that
// introduces a sleep delay of 1 second during each yield.
assert.commandWorked(
coll.getDB().adminCommand({setParameter: 1, internalQueryExecYieldIterations: 10}));
assert.commandWorked(
coll.getDB().adminCommand({setParameter: 1, internalQueryExecYieldPeriodMS: 500}));
assert.commandWorked(coll.getDB().adminCommand({
configureFailPoint: "setYieldAllLocksWait",
namespace: coll.getFullName(),
mode: "alwaysOn",
data: {waitForMillis: 1000}
}));
// Insert 40 documents in the collection, perform a collection scan, and verify that it yields
// about 4 times. Since each group of 10 documents should always be processed in less than 500
// milliseconds, we expect to hit only iteration-based yields for this query, and no
// timing-based yields. 40 documents total divided by 10 documents per yield gives us an
// estimated yield count of 4 yields.
//
// Note also that we have a 1-second sleep delay during each yield, and we expect this delay to
// not change our expectation to hit zero timing-based yields. Timing-based yields only consider
// time spent during query execution since the last yield; since our sleep delay of 1 second is
// not during query execution, it should never count towards our 500 millisecond threshold for a
// timing-based yield (incorrect accounting for timing-based yields was the cause for
// SERVER-21341).
for (var i = 0; i < 40; ++i) {
assert.writeOK(coll.insert({}));
}
var explainRes = coll.find().explain("executionStats");
// We expect 4 yields, but we throw in a fudge factor of 2 for test reliability. We also can
// use "saveState" calls as a proxy for "number of yields" here, because we expect our entire
// result set to be returned in a single batch.
assert.gt(explainRes.executionStats.executionStages.saveState, 4 / 2, tojson(explainRes));
assert.lt(explainRes.executionStats.executionStages.saveState, 4 * 2, tojson(explainRes));
})();
|
