I have the following set – Yarn on top of Hadoop 2.8 and Flink on it and running it on 2-3 nodes.
The problem is that during the running of the job the log files are growing and I’m not able to delete them, cause there is lock on them. I don’t want to stop the job in order to release the lock and delete the logs.
I’ve made some research and most of the people suggests to set some time with the following property yarn.nodemanager.log-aggregation.roll-monitoring-interval-seconds in yarn-site.xml (I set that property in the master and the slaves too).
Unfortunately this doesn’t seems to work – the log is not deleted from the nodes and moved to the master as expected :/ Btw the purpose is to just delete the logs, I don’t need to keep them for more than 24 or 48 hours.
Is there anyone that had similar issue and found a fix for it?