[
https://issues.apache.org/jira/browse/NUTCH-2474?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16283266#comment-16283266
]
Sebastian Nagel commented on NUTCH-2474:
----------------------------------------
The problem is that the CrawlDbStatCombiner is written under the assumption
that is called exactly once between mapper and reducer. However, this
assumption is wrong: "The Combiner may be called 0, 1, or many times on each
key between the mapper and reducer." ([discussion on Hadoop user mailing
list|http://mail-archives.apache.org/mod_mbox/hadoop-common-user/200901.mbox/%[email protected]%3e])
The combiner emits for multiple values of score, fetch time, fetch interval the
minimum, maximum, sum (and with NUTCH-2470 also a T-Digest structure to
estimate quantiles). In distributed mode with a large CrawlDb it may be called
many times and receives min/max/sum as input which is just summed up in the
else-clause. Summing up min/max could be the reason for NUTCH-2297.
> CrawlDbReader -stats fails with ClassCastException
> --------------------------------------------------
>
> Key: NUTCH-2474
> URL: https://issues.apache.org/jira/browse/NUTCH-2474
> Project: Nutch
> Issue Type: Bug
> Components: crawldb
> Affects Versions: 1.14
> Environment: Java 8, distributed mode: Hadoop CDH 5.13.0
> Reporter: Sebastian Nagel
> Priority: Critical
> Fix For: 1.14
>
>
> In distributed mode CrawlDbReader / readdb -stats fails with a
> ClassCastException in the combiner:
> {noformat}
> 17/12/08 04:57:13 INFO mapreduce.Job: Task Id :
> attempt_1512553291624_0022_m_000039_0, Status : FAILED
> Error: java.lang.ClassCastException: org.apache.hadoop.io.FloatWritable
> cannot be cast to org.apache.hadoop.io.LongWritable
> at
> org.apache.nutch.crawl.CrawlDbReader$CrawlDbStatCombiner.reduce(CrawlDbReader.java:296)
> at
> org.apache.nutch.crawl.CrawlDbReader$CrawlDbStatCombiner.reduce(CrawlDbReader.java:222)
> at
> org.apache.hadoop.mapred.Task$OldCombinerRunner.combine(Task.java:1639)
> at
> org.apache.hadoop.mapred.MapTask$MapOutputBuffer.mergeParts(MapTask.java:1946)
> at
> org.apache.hadoop.mapred.MapTask$MapOutputBuffer.flush(MapTask.java:1514)
> at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:466)
> at org.apache.hadoop.mapred.MapTask.run(MapTask.java:343)
> {noformat}
> FloatWritables are used since NUTCH-2470, so that's when this bug was
> introduced.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
