[ https://issues.apache.org/jira/browse/MAPREDUCE-7402?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17832367#comment-17832367 ]
ASF GitHub Bot commented on MAPREDUCE-7402: ------------------------------------------- KeeProMise commented on PR #6622: URL: https://github.com/apache/hadoop/pull/6622#issuecomment-2027895211 > Was gonna hit the merge button, but got curious. > > Earlier if the value was 1, was it an infinite loop always? or in just in a corner case, if in just one corner case & now if we throw an exception, if anyone has value of 1 earlier his jobs would be working but post this change it would fail? > > can we just break the condition for infinite loop if that is the case & preserve the value? or at worst if that is an illegal value, we put a warn log & use the defaul? > > @brumi1024 / @szilard-nemeth any pointers? Hi. The code related to this method has been running for a very long time (since June 13, 2011), and I suspect that setting this value to 1 is meaningless in the actual generation environment. Therefore, this bug should have existed for a long time, but has not been reported yet. Therefore, I believe that if the value is 1, throwing an exception will not have any impact on the existing business. > fix mapreduce.task.io.sort.factor=1 lead to an infinite loop. > ------------------------------------------------------------- > > Key: MAPREDUCE-7402 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-7402 > Project: Hadoop Map/Reduce > Issue Type: Bug > Reporter: Jian Zhang > Priority: Minor > Labels: pull-request-available > Attachments: screenshot-1.png > > > *Describe:* > In the method long computeBytesInMerges(int factor, int inMem), when > factor is set to 1, and the initial numSegments>1, inMem=0, it will cause an > infinite loop of while. Although setting factor to 1 is meaningless, and it > will not be done in actual production, we need to remind users to avoid an > infinite loop due to carelessness. > !screenshot-1.png|width=600,height=400! > *How to reproduce:* > This bug can easily occur by setting > configuration.set("mapreduce.task.io.sort.factor", "1"). -- This message was sent by Atlassian Jira (v8.20.10#820010) --------------------------------------------------------------------- To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org