[
https://issues.apache.org/jira/browse/NUTCH-2518?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16417040#comment-16417040
]
ASF GitHub Bot commented on NUTCH-2518:
---------------------------------------
sebastian-nagel commented on a change in pull request #307: NUTCH-2518 Cleaning
up the file system after a job failure.
URL: https://github.com/apache/nutch/pull/307#discussion_r177675556
##########
File path: src/java/org/apache/nutch/crawl/CrawlDb.java
##########
@@ -129,17 +129,23 @@ public void update(Path crawlDb, Path[] segments,
boolean normalize,
LOG.info("CrawlDb update: Merging segment data into db.");
}
+ FileSystem fs = crawlDb.getFileSystem(getConf());
+ Path outPath = FileOutputFormat.getOutputPath(job);
try {
- int complete = job.waitForCompletion(true)?0:1;
+ boolean success = job.waitForCompletion(true);
+ if (!success) {
+ String message = "Crawl job did not succeed, job status:"
+ + job.getStatus().getState() + ", reason: "
+ + job.getStatus().getFailureInfo();
+ LOG.error(message);
+ cleanupAfterFailure(outPath, lock, fs);
+ throw new RuntimeException(message);
+ }
} catch (IOException | InterruptedException | ClassNotFoundException e) {
- FileSystem fs = crawlDb.getFileSystem(getConf());
- LockUtil.removeLockFile(fs, lock);
- Path outPath = FileOutputFormat.getOutputPath(job);
- if (fs.exists(outPath))
- fs.delete(outPath, true);
+ LOG.error("Crawl job failed ", e);
Review comment:
+1
The CrawlDb cleanup method looks same as that from Injector: maybe move it
to CrawlDb - there are already methods to lock and install the CrawlDb shared
by various jobs (inject, update, merge, generate).
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> Must check return value of job.waitForCompletion()
> --------------------------------------------------
>
> Key: NUTCH-2518
> URL: https://issues.apache.org/jira/browse/NUTCH-2518
> Project: Nutch
> Issue Type: Bug
> Components: crawldb, fetcher, generator, hostdb, linkdb
> Affects Versions: 1.15
> Reporter: Sebastian Nagel
> Assignee: Kenneth McFarland
> Priority: Blocker
> Fix For: 1.15
>
>
> The return value of job.waitForCompletion() of the new MapReduce API
> (NUTCH-2375) must always be checked. If it's not true, the job has been
> failed or killed. Accordingly, the program
> - should not proceed with further jobs/steps
> - must clean-up temporary data, unlock CrawlDB, etc.
> - exit with non-zero exit value, so that scripts running the crawl workflow
> can handle the failure
> Cf. NUTCH-2076, NUTCH-2442, [NUTCH-2375 PR
> #221|https://github.com/apache/nutch/pull/221#issuecomment-332941883].
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
