其他分享
首页 > 其他分享> > Building a Sync Engin

Building a Sync Engin

作者：互联网

内容来自：https://www.grouparoo.com/blog/building-a-sync-engine
内容主要介绍了如何开发一个同步引擎，没有太多高深的，主要是基于了变动的时间戳以及水印算法

简单说明

  
ALTER TABLE users ADD COLUMN mysql_updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP;
  
const watermark = await getWatermark();
let rows;
if (!watermark) {
  // first time we've ever sync'd - get all rows
  rows = await User.findAll();
} else {
  rows = await User.findAll({
    // otherwise, use watermark
    where: {
      updatedAt: {
        [Op.gt]: watermark, // WHERE updatedAt > {watermark}
      },
    },
    order: [["updatedAt", "ASC"]],
  });
}
  
if (rows && rows.length > 0) {
  for (const row of rows) {
    await processRow(row);
  }
  
  const newWatermark = new Date(); // set to now
  await setWatermark(newWatermark); // for next time
}
return true; // done!

问题：数据只能增长，数据同步的时候可能会变动，数据可能会重复处理，时间可能会不一致（服务器时间戳，，，）
修复,时间问题，基于db 时间

  
const watermark = await getWatermark();
let rows;
if (!watermark) {
  // first time we've ever sync'd - get all rows
  rows = await User.findAll();
} else {
  rows = await User.findAll({
    // otherwise, use watermark
    where: {
      updatedAt: {
        [Op.gte]: watermark, // WHERE updatedAt >= {watermark}
      },
    },
    order: [["updatedAt", "ASC"]],
  });
}
  
if (rows && rows.length > 0) {
  for (const row of rows) {
    await processRow(row);
  }
  
  const newWatermark = rows[rows.length - 1].updatedAt;
  await setWatermark(newWatermark); // for next time
}
return true; // done!
  
// using node and sequelize
const saved = await getWatermark();
const watermark = saved ? saved.watermark : null;
const oldOffset = saved ? saved.offset || 0 : null;
const sqlOptions = {
  limit: batchSize,
  offset: oldOffset,
  order: [["updatedAt", "ASC"]],
};
  
if (watermark) {
  sqlOptions.where = {
    updatedAt: {
      [Op.gte]: watermark, // WHERE updatedAt >= {watermark}
    },
  };
}
  
const rows = await User.findAll(sqlOptions);
if (!rows || rows.length === 0) {
  return true;
} else {
  for (const row of rows) {
    await processRow(row);
  }
  
  const done = rows.length < batchSize; // is there more to be done?
  const lastTime = rows[rows.length - 1].updatedAt.getTime();
  let newOffset = 0;
  if (!done && watermark === lastTime) {
    // the last one was the same as the first, need to use offset
    newOffset = oldOffset + batchSize;
  }
  
  await setWatermark({ watermark: lastTime, offset: newOffset });
  return done;
}

说明

以上方法还是值得参考学习的，尽管有时我们是不能直接使用的，但是还是很不错的实践，cdc，except 有时可能会是一个其他的选择

参考资

https://www.grouparoo.com/blog/building-a-sync-engine
https://github.com/grouparoo/sync-engine-example

标签：Building,rows,const,watermark,await,Sync,updatedAt,Engin,row
来源： https://www.cnblogs.com/rongfengliang/p/15859650.html