java - When are Hadoop counters persisted? -
i have simple mapreduce job uses hadoop counters count unique ids
in reduce phase of job.
are counter changes persisted when increment java code, or after successful execution of task?
the reason i'm wondering find out if should disable speculative execution in job, cause double counting unless counter changes happen after successful execution of task. , then, there race conditions if identical tasks succeed around same time?
Comments
Post a Comment