聊聊maxwell的BinlogConnectorEventListener
来源:SegmentFault
2023-02-24 16:31:57
0浏览
收藏
你在学习数据库相关的知识吗?本文《聊聊maxwell的BinlogConnectorEventListener》,主要介绍的内容就涉及到MySQL,如果你想提升自己的开发能力,就不要错过这篇文章,大家要知道编程理论基础和实战操作都是不可或缺的哦!
序
本文主要研究一下maxwell的BinlogConnectorEventListener
BinlogConnectorEventListener
maxwell-1.25.1/src/main/java/com/zendesk/maxwell/replication/BinlogConnectorEventListener.java
class BinlogConnectorEventListener implements BinaryLogClient.EventListener {
private static final Logger LOGGER = LoggerFactory.getLogger(BinlogConnectorEventListener.class);
private final BlockingQueue queue;
private final Timer queueTimer;
protected final AtomicBoolean mustStop = new AtomicBoolean(false);
private final BinaryLogClient client;
private final MaxwellOutputConfig outputConfig;
private long replicationLag;
private String gtid;
public BinlogConnectorEventListener(
BinaryLogClient client,
BlockingQueue q,
Metrics metrics,
MaxwellOutputConfig outputConfig
) {
this.client = client;
this.queue = q;
this.queueTimer = metrics.getRegistry().timer(metrics.metricName("replication", "queue", "time"));
this.outputConfig = outputConfig;
final BinlogConnectorEventListener self = this;
metrics.register(metrics.metricName("replication", "lag"), (Gauge) () -> self.replicationLag);
}
public void stop() {
mustStop.set(true);
}
@Override
public void onEvent(Event event) {
long eventSeenAt = 0;
boolean trackMetrics = false;
if (event.getHeader().getEventType() == EventType.GTID) {
gtid = ((GtidEventData)event.getData()).getGtid();
}
BinlogConnectorEvent ep = new BinlogConnectorEvent(event, client.getBinlogFilename(), client.getGtidSet(), gtid, outputConfig);
if (ep.isCommitEvent()) {
trackMetrics = true;
eventSeenAt = System.currentTimeMillis();
replicationLag = eventSeenAt - event.getHeader().getTimestamp();
}
while (mustStop.get() != true) {
try {
if ( queue.offer(ep, 100, TimeUnit.MILLISECONDS ) ) {
break;
}
} catch (InterruptedException e) {
return;
}
}
if (trackMetrics) {
queueTimer.update(System.currentTimeMillis() - eventSeenAt, TimeUnit.MILLISECONDS);
}
}
}
- BinlogConnectorEventListener实现了BinaryLogClient.EventListener接口,其onEvent方法创建BinlogConnectorEvent,然后将该event放入queue中
BinlogConnectorEvent
maxwell-1.25.1/src/main/java/com/zendesk/maxwell/replication/BinlogConnectorEvent.java
public class BinlogConnectorEvent {
public static final String BEGIN = "BEGIN";
public static final String COMMIT = "COMMIT";
public static final String SAVEPOINT = "SAVEPOINT";
private final MaxwellOutputConfig outputConfig;
private BinlogPosition position;
private BinlogPosition nextPosition;
private final Event event;
private final String gtidSetStr;
private final String gtid;
public BinlogConnectorEvent(Event event, String filename, String gtidSetStr, String gtid, MaxwellOutputConfig outputConfig) {
this.event = event;
this.gtidSetStr = gtidSetStr;
this.gtid = gtid;
EventHeaderV4 hV4 = (EventHeaderV4) event.getHeader();
this.nextPosition = new BinlogPosition(gtidSetStr, gtid, hV4.getNextPosition(), filename);
this.position = new BinlogPosition(gtidSetStr, gtid, hV4.getPosition(), filename);
this.outputConfig = outputConfig;
}
public Event getEvent() {
return event;
}
public WriteRowsEventData writeRowsData() {
return (WriteRowsEventData) event.getData();
}
public UpdateRowsEventData updateRowsData() {
return (UpdateRowsEventData) event.getData();
}
public DeleteRowsEventData deleteRowsData() {
return (DeleteRowsEventData) event.getData();
}
public QueryEventData queryData() {
return (QueryEventData) event.getData();
}
public XidEventData xidData() {
return (XidEventData) event.getData();
}
public TableMapEventData tableMapData() {
return (TableMapEventData) event.getData();
}
public BinlogPosition getPosition() {
return position;
}
public BinlogPosition getNextPosition() {
return nextPosition;
}
public EventType getType() {
return event.getHeader().getEventType();
}
public Long getTableID() {
EventData data = event.getData();
switch ( event.getHeader().getEventType() ) {
case EXT_WRITE_ROWS:
case WRITE_ROWS:
return ((WriteRowsEventData) data).getTableId();
case EXT_UPDATE_ROWS:
case UPDATE_ROWS:
return ((UpdateRowsEventData) data).getTableId();
case EXT_DELETE_ROWS:
case DELETE_ROWS:
return ((DeleteRowsEventData) data).getTableId();
case TABLE_MAP:
return ((TableMapEventData) data).getTableId();
}
return null;
}
public boolean isCommitEvent() {
EventType eventType = getType();
if (eventType == EventType.XID) {
return true;
} else if (eventType == EventType.QUERY) {
// MyISAM will output a "COMMIT" QUERY_EVENT instead of a XID_EVENT.
// There's no transaction ID but we can still set "commit: true"
return COMMIT.equals(queryData().getSql());
}
return false;
}
private void writeData(Table table, RowMap row, Serializable[] data, BitSet includedColumns) {
int dataIdx = 0, colIdx = 0;
for ( ColumnDef cd : table.getColumnList() ) {
if ( includedColumns.get(colIdx) ) {
Object json = null;
if ( data[dataIdx] != null ) {
json = cd.asJSON(data[dataIdx], outputConfig);
}
row.putData(cd.getName(), json);
dataIdx++;
}
colIdx++;
}
}
private void writeOldData(Table table, RowMap row, Serializable[] oldData, BitSet oldIncludedColumns) {
int dataIdx = 0, colIdx = 0;
for ( ColumnDef cd : table.getColumnList() ) {
if ( oldIncludedColumns.get(colIdx) ) {
Object json = null;
if ( oldData[dataIdx] != null ) {
json = cd.asJSON(oldData[dataIdx], outputConfig);
}
if (!row.hasData(cd.getName())) {
/*
If we find a column in the BEFORE image that's *not* present in the AFTER image,
we're running in binlog_row_image = MINIMAL. In this case, the BEFORE image acts
as a sort of WHERE clause to update rows with the new values (present in the AFTER image),
In this case we should put what's in the "before" image into the "data" section, not the "old".
*/
row.putData(cd.getName(), json);
} else {
if (!Objects.equals(row.getData(cd.getName()), json)) {
row.putOldData(cd.getName(), json);
}
}
dataIdx++;
}
colIdx++;
}
}
private RowMap buildRowMap(String type, Position position, Position nextPosition, Serializable[] data, Table table, BitSet includedColumns, String rowQuery) {
RowMap map = new RowMap(
type,
table.getDatabase(),
table.getName(),
event.getHeader().getTimestamp(),
table.getPKList(),
position,
nextPosition,
rowQuery
);
writeData(table, map, data, includedColumns);
return map;
}
public List jsonMaps(Table table, long lastHeartbeatRead, String rowQuery) {
ArrayList list = new ArrayList();
Position position = Position.valueOf(this.position, lastHeartbeatRead);
Position nextPosition = Position.valueOf(this.nextPosition, lastHeartbeatRead);
switch ( getType() ) {
case WRITE_ROWS:
case EXT_WRITE_ROWS:
for ( Serializable[] data : writeRowsData().getRows() ) {
list.add(buildRowMap("insert", position, nextPosition, data, table, writeRowsData().getIncludedColumns(), rowQuery));
}
break;
case DELETE_ROWS:
case EXT_DELETE_ROWS:
for ( Serializable[] data : deleteRowsData().getRows() ) {
list.add(buildRowMap("delete", position, nextPosition, data, table, deleteRowsData().getIncludedColumns(), rowQuery));
}
break;
case UPDATE_ROWS:
case EXT_UPDATE_ROWS:
for ( Map.Entry e : updateRowsData().getRows() ) {
Serializable[] data = e.getValue();
Serializable[] oldData = e.getKey();
RowMap r = buildRowMap("update", position, nextPosition, data, table, updateRowsData().getIncludedColumns(), rowQuery);
writeOldData(table, r, oldData, updateRowsData().getIncludedColumnsBeforeUpdate());
list.add(r);
}
break;
}
return list;
}
}
- BinlogConnectorEvent提供了jsonMaps方法,该方法根据event.getHeader().getEventType()类型来做不同处理,主要通过buildRowMap、writeOldData这几个方法来实现;buildRowMap方法返回RowMap类型,writeOldData方法主要判断是否putOldData
BinlogConnectorReplicator
maxwell-1.25.1/src/main/java/com/zendesk/maxwell/replication/BinlogConnectorReplicator.java
public class BinlogConnectorReplicator extends RunLoopProcess implements Replicator {
//......
public RowMap getRow() throws Exception {
BinlogConnectorEvent event;
if ( stopOnEOF && hitEOF )
return null;
if ( !replicatorStarted )
throw new ReplicatorNotReadyException("replicator not started!");
while (true) {
if (rowBuffer != null && !rowBuffer.isEmpty()) {
RowMap row = rowBuffer.removeFirst();
if ( row != null && isMaxwellRow(row) && row.getTable().equals("heartbeats") )
return processHeartbeats(row);
else
return row;
}
event = pollEvent();
if (event == null) {
if ( stopOnEOF ) {
if ( client.isConnected() )
continue;
else
return null;
} else {
try {
ensureReplicatorThread();
} catch ( ClientReconnectedException e ) {}
return null;
}
}
switch (event.getType()) {
case WRITE_ROWS:
case EXT_WRITE_ROWS:
case UPDATE_ROWS:
case EXT_UPDATE_ROWS:
case DELETE_ROWS:
case EXT_DELETE_ROWS:
LOGGER.warn("Started replication stream inside a transaction. This shouldn't normally happen.");
LOGGER.warn("Assuming new transaction at unexpected event:" + event);
queue.offerFirst(event);
rowBuffer = getTransactionRows(event);
break;
case TABLE_MAP:
TableMapEventData data = event.tableMapData();
tableCache.processEvent(getSchema(), this.filter, data.getTableId(), data.getDatabase(), data.getTable());
break;
case QUERY:
QueryEventData qe = event.queryData();
String sql = qe.getSql();
if (BinlogConnectorEvent.BEGIN.equals(sql)) {
try {
rowBuffer = getTransactionRows(event);
} catch ( ClientReconnectedException e ) {
// rowBuffer should already be empty by the time we get to this switch
// statement, but we null it for clarity
rowBuffer = null;
break;
}
rowBuffer.setServerId(event.getEvent().getHeader().getServerId());
rowBuffer.setThreadId(qe.getThreadId());
rowBuffer.setSchemaId(getSchemaId());
} else {
processQueryEvent(event);
}
break;
case ROTATE:
tableCache.clear();
if ( stopOnEOF && event.getPosition().getOffset() > 0 ) {
this.binlogEventListener.mustStop.set(true);
this.client.disconnect();
this.hitEOF = true;
return null;
}
break;
default:
break;
}
}
}
protected BinlogConnectorEvent pollEvent() throws InterruptedException {
return queue.poll(100, TimeUnit.MILLISECONDS);
}
//......
}
- BinlogConnectorReplicator提供了getRow方法,它先从rowBuffer取row,取不到则通过pollEvent方法从queue中取event,之后根据不同的event.getType()来做不同处理,有的会通过getTransactionRows(event)来给rowBuffer赋值
小结
BinlogConnectorEventListener实现了BinaryLogClient.EventListener接口,其onEvent方法创建BinlogConnectorEvent,然后将该event放入queue中;BinlogConnectorReplicator提供了getRow方法,它会通过pollEvent从queue中拉取BinlogConnectorEvent,解析为RowMap
doc
今天带大家了解了MySQL的相关知识,希望对你有所帮助;关于数据库的技术知识我们会一点点深入介绍,欢迎大家关注golang学习网公众号,一起学习编程~
版本声明
本文转载于:SegmentFault 如有侵犯,请联系study_golang@163.com删除
聊聊BinlogConnectorReplicator的work
- 上一篇
- 聊聊BinlogConnectorReplicator的work
- 下一篇
- 推荐一款优秀的前后端开发框架Open-Admin
查看更多
最新文章
-
- 数据库 · MySQL | 1天前 |
- MySQL 事务隔离级别下间隙锁影响范围的分析方法
- 390浏览 收藏
-
- 数据库 · MySQL | 4天前 |
- MySQL EXPLAIN ANALYZE定位排序临时表的排查方法
- 434浏览 收藏
-
- 数据库 · MySQL | 4天前 | MySQL · JSON ·
- MySQL JSON多值索引处理数组成员检索的设计要点
- 404浏览 收藏
-
- 数据库 · MySQL | 4天前 | MySQL · 数据库 · mysql SQL JSON JSON_TABLE
- MySQL JSON_TABLE为缺失字段提供默认值的映射方法
- 445浏览 收藏
-
- 数据库 · MySQL | 4天前 |
- MySQL CTE拆分多阶段聚合查询的维护方法
- 418浏览 收藏
-
- 数据库 · MySQL | 4天前 | MySQL · 数据库 · mysql 窗口函数 ROW_NUMBER RANK DENSE_RANK
- MySQL窗口函数按分组取排名前N条的查询设计
- 465浏览 收藏
查看更多
课程推荐
-
- 前端进阶之JavaScript设计模式
- 设计模式是开发人员在软件开发过程中面临一般问题时的解决方案,代表了最佳的实践。本课程的主打内容包括JS常见设计模式以及具体应用场景,打造一站式知识长龙服务,适合有JS基础的同学学习。
- 543次学习
-
- GO语言核心编程课程
- 本课程采用真实案例,全面具体可落地,从理论到实践,一步一步将GO核心编程技术、编程思想、底层实现融会贯通,使学习者贴近时代脉搏,做IT互联网时代的弄潮儿。
- 516次学习
-
- 简单聊聊mysql8与网络通信
- 如有问题加微信:Le-studyg;在课程中,我们将首先介绍MySQL8的新特性,包括性能优化、安全增强、新数据类型等,帮助学生快速熟悉MySQL8的最新功能。接着,我们将深入解析MySQL的网络通信机制,包括协议、连接管理、数据传输等,让
- 500次学习
-
- JavaScript正则表达式基础与实战
- 在任何一门编程语言中,正则表达式,都是一项重要的知识,它提供了高效的字符串匹配与捕获机制,可以极大的简化程序设计。
- 487次学习
-
- 从零制作响应式网站—Grid布局
- 本系列教程将展示从零制作一个假想的网络科技公司官网,分为导航,轮播,关于我们,成功案例,服务流程,团队介绍,数据部分,公司动态,底部信息等内容区块。网站整体采用CSSGrid布局,支持响应式,有流畅过渡和展现动画。
- 485次学习
查看更多
AI推荐
-
- PubMedQA
- 深入了解PubMedQA生物医学问答数据集,涵盖其核心功能、使用方法及在临床决策、药物研发等场景的应用,助力提升NLP模型性能。
- 198次使用
-
- H2O EvalGPT
- H2O EvalGPT是H2O.ai推出的开源LLM评估平台,提供详细的大模型性能排行榜、行业特定基准测试及A/B测试功能,助您快速选择最适合项目的高性能大语言模型。
- 255次使用
-
- LMArena
- LMArena是加州大学伯克利分校推出的AI模型匿名评测平台。通过盲测投票机制,用户可对比不同大模型回答并生成实时排行榜,助力开发者优化模型及用户选择最佳AI工具。
- 209次使用
-
- HELM
- 深入了解斯坦福推出的HELM(Holistic Evaluation of Language Models)大模型评测体系。本文解析其核心功能、安装配置步骤及应用场景,涵盖准确性、公平性、鲁棒性等多维度指标,助力开发者全面优化语言模型性能。
- 196次使用
-
- CMMLU
- 深入了解CMMLU中文评估基准,涵盖67个学科主题,提供数据集下载、Zero-shot/Five-shot评估方法及排行榜,助力优化中文语言模型性能。
- 185次使用
查看更多
相关文章
-
- MySQL 明明加了索引,为什么查询还是很慢?先查这 6 个点
- 2026-06-27 374浏览
-
- golang MySQL实现对数据库表存储获取操作示例
- 2022-12-22 499浏览
-
- golang 基于 mysql 简单实现分布式读写锁
- 2023-01-07 384浏览
-
- 详解如何利用GORM实现MySQL事务
- 2023-01-07 184浏览
-
- Go语言实现操作MySQL的基础知识总结
- 2023-01-23 265浏览

