0
0
mirror of https://github.com/mongodb/mongo.git synced 2024-11-25 09:19:32 +01:00
mongodb/jstests/noPassthrough/readConcern_atClusterTime_snapshot_selection.js

142 lines
6.0 KiB
JavaScript
Raw Normal View History

// Test that 'atClusterTime' is used to select the snapshot for reads. We wait for 'atClusterTime'
// to be majority committed. If 'atClusterTime' is older than the oldest available snapshot, the
// error code SnapshotTooOld is returned.
//
// @tags: [requires_replication]
(function() {
"use strict";
load("jstests/libs/write_concern_util.js"); // For stopServerReplication.
const dbName = "test";
const collName = "coll";
const rst = new ReplSetTest({nodes: 3, settings: {chainingAllowed: false}});
rst.startSet();
rst.initiate();
const primarySession =
rst.getPrimary().getDB(dbName).getMongo().startSession({causalConsistency: false});
const primaryDB = primarySession.getDatabase(dbName);
let primaryTxnNumber = 0;
if (!assert.commandWorked(primaryDB.serverStatus()).storageEngine.supportsSnapshotReadConcern) {
rst.stopSet();
return;
}
const secondaryConns = rst.getSecondaries();
const secondaryConn0 = secondaryConns[0];
const secondaryConn1 = secondaryConns[1];
const secondarySession =
secondaryConn0.getDB(dbName).getMongo().startSession({causalConsistency: false});
const secondaryDB0 = secondarySession.getDatabase(dbName);
let secondaryTxnNumber = 0;
// Create the collection and insert one document. Get the op time of the write.
let res = assert.commandWorked(primaryDB.runCommand(
{insert: collName, documents: [{_id: "before"}], writeConcern: {w: "majority"}}));
let clusterTimePrimaryBefore;
// Wait for the majority commit point on 'secondaryDB0' to include the {_id: "before"} write.
assert.soonNoExcept(function() {
// Without a consistent stream of writes, secondary majority reads are not guaranteed
// to complete, since the commit point being stale is not sufficient to establish a sync
// source.
// TODO (SERVER-33248): Remove this write and increase the maxTimeMS on the read.
res = assert.commandWorked(primaryDB.runCommand(
{insert: "otherColl", documents: [{a: 1}], writeConcern: {w: "majority"}}));
assert(res.hasOwnProperty("opTime"), tojson(res));
assert(res.opTime.hasOwnProperty("ts"), tojson(res));
clusterTimePrimaryBefore = res.opTime.ts;
return assert
.commandWorked(secondaryDB0.runCommand(
{find: collName, readConcern: {level: "majority"}, maxTimeMS: 10000}))
.cursor.firstBatch.length === 1;
});
// Stop replication on both secondaries.
stopServerReplication(secondaryConn0);
stopServerReplication(secondaryConn1);
// Perform write and get the op time of the write.
res =
assert.commandWorked(primaryDB.runCommand({insert: collName, documents: [{_id: "after"}]}));
assert(res.hasOwnProperty("opTime"), tojson(res));
assert(res.opTime.hasOwnProperty("ts"), tojson(res));
let clusterTimeAfter = res.opTime.ts;
// A read on the primary at the old cluster time should not include the write.
res = assert.commandWorked(primaryDB.runCommand({
find: collName,
readConcern: {level: "snapshot", atClusterTime: clusterTimePrimaryBefore},
txnNumber: NumberLong(primaryTxnNumber++)
}));
assert.eq(res.cursor.firstBatch.length, 1, printjson(res));
assert.eq(res.cursor.firstBatch[0]._id, "before", printjson(res));
// A read on the primary at the new cluster time should time out waiting for the cluster time to
// be majority committed.
assert.commandFailedWithCode(primaryDB.runCommand({
find: collName,
readConcern: {level: "snapshot", atClusterTime: clusterTimeAfter},
maxTimeMS: 1000,
txnNumber: NumberLong(primaryTxnNumber++)
}),
ErrorCodes.ExceededTimeLimit);
// Restart replication on one of the secondaries.
restartServerReplication(secondaryConn1);
// A read on the primary at the new cluster time now succeeds.
res = assert.commandWorked(primaryDB.runCommand({
find: collName,
readConcern: {level: "snapshot", atClusterTime: clusterTimeAfter},
txnNumber: NumberLong(primaryTxnNumber++)
}));
assert.eq(res.cursor.firstBatch.length, 2, printjson(res));
// A read on the lagged secondary at its view of the majority cluster time should not include
// the write.
const clusterTimeSecondaryBefore = rst.getReadConcernMajorityOpTimeOrThrow(secondaryConn0).ts;
res = assert.commandWorked(secondaryDB0.runCommand({
find: collName,
readConcern: {level: "snapshot", atClusterTime: clusterTimeSecondaryBefore},
txnNumber: NumberLong(secondaryTxnNumber++)
}));
assert.eq(res.cursor.firstBatch.length, 1, printjson(res));
assert.eq(res.cursor.firstBatch[0]._id, "before", printjson(res));
// A read on the lagged secondary at the new cluster time should time out waiting for an op at
// that cluster time.
assert.commandFailedWithCode(secondaryDB0.runCommand({
find: collName,
readConcern: {level: "snapshot", atClusterTime: clusterTimeAfter},
maxTimeMS: 1000,
txnNumber: NumberLong(secondaryTxnNumber++)
}),
ErrorCodes.ExceededTimeLimit);
// Restart replication on the lagged secondary.
restartServerReplication(secondaryConn0);
// A read on the secondary at the new cluster time now succeeds.
res = assert.commandWorked(secondaryDB0.runCommand({
find: collName,
readConcern: {level: "snapshot", atClusterTime: clusterTimeAfter},
txnNumber: NumberLong(secondaryTxnNumber++)
}));
assert.eq(res.cursor.firstBatch.length, 2, printjson(res));
// A read at a time that is too old fails.
assert.commandFailedWithCode(primaryDB.runCommand({
find: collName,
readConcern: {level: "snapshot", atClusterTime: Timestamp(1, 1)},
txnNumber: NumberLong(primaryTxnNumber++)
}),
ErrorCodes.SnapshotTooOld);
rst.stopSet();
}());