cascading - how to avoid filling up hadoop logs on nodes? -
when our cascading jobs encounter error in data, throw various exceptions… these end in logs, , if logs fill up, cluster stops working. have config file edited/configured avoid such scenarios?
we using mapr 3.1.0, , looking way limit log use (syslogs/userlogs), without using centralized logging, without adjusting logging level, , less bothered whether keeps first n bytes, or last n bytes of logs , discords remain part.
we don't care logs, , need first (or last) few megs figure out went wrong. don't wan't use centralized logging, because don't want keep logs/ don't care spend perf overhead of replicating them. also, correct me if i'm wrong: user_log.retain-size, has issues when jvm re-use used.
any clue/answer appreciated !!
thanks,
srinivas
this should in different stackexchange site it's more of devops question programming question.
anyway, need devops setup logrotate , configure according needs. or edit log4j xml files on cluster change way logging done.
Comments
Post a Comment