http://caols.diandian.com/post/2012-06-05/40028026285
http://blog.csdn.net/witsmakemen/article/details/7799546 这个是相关的代码的分析
昨天看错了,实际上对于爬取成功的url,在update()阶段,程序会将url的FetchTime+FetchInterval作为最终的下次FetchTime,这个FetchTime已经不再代表网页成功Fetch的时间,而是作为下次Fetch的时间,如果在小于新的FetchTime的时间内对该url进行爬去,程序将会过滤掉该url。
在CrawlDbReducer中的reduce函数:
[java] view plaincopy
case CrawlDatum.STATUS_FETCH_SUCCESS: // succesful fetch
case CrawlDatum.STATUS_FETCH_REDIR_TEMP: // successful fetch, redirected
case CrawlDatum.STATUS_FETCH_REDIR_PERM:
case CrawlDatum.STATUS_FETCH_NOTMODIFIED: // successful fetch, notmodified
// determine the modification status
int modified = FetchSchedule.STATUS_UNKNOWN;
if (fetch.getStatus() == CrawlDatum.STATUS_FETCH_NOTMODIFIED) {
modified = FetchSchedule.STATUS_NOTMODIFIED;
} else {
if (oldSet && old.getSignature() != null && signature != null) {
if (SignatureComparator._compare(old.getSignature(), signature) != 0) {
modified = FetchSchedule.STATUS_MODIFIED;
} else {
modified = FetchSchedule.STATUS_NOTMODIFIED;
}
}
}
// set the schedule
System.err.println("1:result.fetchtime="+result.getFetchTime());
result = schedule.setFetchSchedule((Text)key, result, prevFetchTime,
prevModifiedTime, fetch.getFetchTime(), fetch.getModifiedTime(), modified);
// set the result status and signature
System.err.println("2:result.fetchtime="+result.getFetchTime());
if (modified == FetchSchedule.STATUS_NOTMODIFIED) {
result.setStatus(CrawlDatum.STATUS_DB_NOTMODIFIED);
if (oldSet) result.setSignature(old.getSignature());
} else {
switch (fetch.getStatus()) {
case CrawlDatum.STATUS_FETCH_SUCCESS:
result.setStatus(CrawlDatum.STATUS_DB_FETCHED);
break;
case CrawlDatum.STATUS_FETCH_REDIR_PERM:
result.setStatus(CrawlDatum.STATUS_DB_REDIR_PERM);
break;
case CrawlDatum.STATUS_FETCH_REDIR_TEMP:
result.setStatus(CrawlDatum.STATUS_DB_REDIR_TEMP);
break;
default:
LOG.warn("Unexpected status: " + fetch.getStatus() + " resetting to old status.");
if (oldSet) result.setStatus(old.getStatus());
else result.setStatus(CrawlDatum.STATUS_DB_UNFETCHED);
}
result.setSignature(signature);
if (metaFromParse != null) {
for (Entry<Writable, Writable> e : metaFromParse.entrySet()) {
result.getMetaData().put(e.getKey(), e.getValue());
}
}
}
// if fetchInterval is larger than the system-wide maximum, trigger
// an unconditional recrawl. This prevents the page to be stuck at
// NOTMODIFIED state, when the old fetched copy was already removed with
// old segments.
if (maxInterval < result.getFetchInterval())
result = schedule.forceRefetch((Text)key, result, false);
break;
通过跟踪打印result的FetchTime值的情况,可以发现,程序在调用schedule.setFetchSchedule()函数之后,值FetchTime的值发生了变化,所以可以肯定是该函数改变了当前url的状态类CrawlDatum的FetchTime状态。
CrawlDbReducer类中,调用的FetchSchedule的扩展为DefaultFetchSchedule类,他的源代码:
[java] view plaincopy
public class DefaultFetchSchedule extends AbstractFetchSchedule {
@Override
public CrawlDatum setFetchSchedule(Text url, CrawlDatum datum,
long prevFetchTime, long prevModifiedTime,
long fetchTime, long modifiedTime, int state) {
// System.err.println("+++++++++++++++++++555555555555555+++++++++++++>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>");
datum = super.setFetchSchedule(url, datum, prevFetchTime, prevModifiedTime,
fetchTime, modifiedTime, state);
if (datum.getFetchInterval() == 0 ) {
datum.setFetchInterval(defaultInterval);
}
datum.setFetchTime(fetchTime + (long)datum.getFetchInterval() * 1000);
datum.setModifiedTime(modifiedTime);
return datum;
}
}
可以看到该类中,只有一个方法setFetchSchedule(),该函数最终将datum的FetchTime的值设置为 datum.setFetchTime(fetchTime + (long)datum.getFetchInterval() * 1000);
主要参考资料
1 http://caols.diandian.com/post/2012-06-05/40028026285
2 http://blog.csdn.net/witsmakemen/article/details/7799546
3 AdaptiveFetchSchedule 类的文档阅读
关键属性解析:Interval
1 InjectMapper中 interval = jobConf.getInt("db.fetch.interval.default", 2592000);
2 int customInterval = interval; customInterval = Integer.parseInt(metavalue); 或者从URL的interval中设置
3 CrawlDatum datum = new CrawlDatum(CrawlDatum.STATUS_INJECTED,customInterval); 被放入CrawlDataum
4 CrawlDb update() 中的 CrawlDbReducer 中的
result = schedule.setFetchSchedule((Text)key, result, prevFetchTime,prevModifiedTime, fetch.getFetchTime(), fetch.getModifiedTime(), modified);
5 schedule默认是调用 DefaultFetchSchedule 这里可以设置下一次采集的日期,方法是:设置索引页过期时间1800 (半小时) 设置 内容页过期时间(7776000 90天)
<property>
<name>db.fetch.interval.default</name>
<value>86400</value>
<description>this usage change to be the interval of index page
this property has wasted by wqj 2013-1-18 The default number of seconds between re-fetches of a page (30 days).
</description>
</property>
<property>
<name>db.fetch.interval.content</name>
<value>7776000</value>
<description>interval for content page ,default is (90 days).
</description>
</property>
/* modify time :2013-01-18 author wqj */
Configuration confCurrent=super.getConf();
int interval_index= confCurrent.getInt("db.fetch.interval.default", 86400);//默认为24小时
int interval_content=confCurrent.getInt("db.fetch.interval.content", 7776000);//默认为90天
String regi = datum.getMetaData().get(new Text("regi")).toString();
if (url.toString().matches(regi)) {
datum.setFetchInterval(interval_index);
} else {
datum.setFetchInterval(interval_content);
}
datum.setFetchTime(fetchTime + (long) datum.getFetchInterval()
* 1000);
datum.setModifiedTime(modifiedTime);
return datum;
采集时间 2013-1-18 14 :30 左右
http://glcx.moc.gov.cn/CsckManageAction/cxckMoreInfo.do?byName=lkbs_newRoad&page=9 Version: 7
Status: 2 (db_fetched)
Fetch time: Fri Jan 18 14:59:53 CST 2013 这里的采集月份依然是1月份 但是也加了 半小时了
Modified time: Thu Jan 01 08:00:00 CST 1970
Retries since fetch: 0
Retry interval: 1800 seconds (0 days)
http://glcx.moc.gov.cn/CsckManageAction/cxckInformationAction.do?infoId=8a8181d532b43eb40132b9076dd20253 Version: 7
Status: 2 (db_fetched)
Fetch time: Thu Apr 18 14:26:26 CST 2013 采集月份是2013 1月份 这里 fetchtime 变为了 4月份 说明这个时间是加上 Retry interval的
Modified time: Thu Jan 01 08:00:00 CST 1970
Retries since fetch: 0
Retry interval: 7776000 seconds (90 days)
Score: 0.009259259
Signature: e901870589199bb918c45c9c9fad0782
通过上述可以看出
fetchTime 事实上在上一次 采集的时候 已经计算好了
下一次将直接从 CrawlerDb中 读取<当前时间的URL 进行采集
引申:
如果想查看内容页 更新 比方一段时间后内容更改了,但是啊 我们那边是用 URL 布隆过滤器去重的,即使有更新也不会反应在数据库里啊
是不是 RegexDumpParser 在这里 观察一下 更新的网页的 DUMP文件是否有什么特殊标记,如果是 有更新内容的文章 则改为UPDATE呢