[ https://issues.apache.org/jira/browse/HADOOP-1984?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Amar Kamat updated HADOOP-1984: ------------------------------- Status: Patch Available (was: Open) Taking into consideration Arun's comment, hard coding the computations for base=2. The way it works now is as follows 1. the first notification is sent to the jobtracker after {{max-backoff}} time. The backoff values are ({{4, 8, 16, 32 ..... k}}) such that {{4 + 8 + 16 + 32 .... k ~ max-backoff}}, {{max-backoff}} can be set using {{mapred.reduce.copy.backoff}} [default is 300 sec]. 2. subsequent notifications are also sent after {{max-backoff}} time but the backoff values now are ({{max-backoff/2, max-backoff/2}}) , i.e 2 retries 3. at the max the reducer waits for {{3 * max-backoff}} before the maps get re-executed. Comments ? > some reducer stuck at copy phase and progress extremely slowly > -------------------------------------------------------------- > > Key: HADOOP-1984 > URL: https://issues.apache.org/jira/browse/HADOOP-1984 > Project: Hadoop > Issue Type: Bug > Components: mapred > Affects Versions: 0.16.0 > Reporter: Runping Qi > Assignee: Amar Kamat > Priority: Critical > Fix For: 0.16.0 > > Attachments: HADOOP-1984-simple.patch, HADOOP-1984-simple.patch, > HADOOP-1984-simple.patch, HADOOP-1984.patch > > > In many cases, some reducers got stuck at copy phase, progressing extremely > slowly. > The entire cluster seems doing nothing. This causes a very bad long tails of > otherwise well tuned map/red jobs. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.