Previously only he adjusted the DFS storage directory to the maximum quota of directories, others were not processed (that is, in the default/directory, and the storage quota for this directory is only 50G)
Run a week less time, cluster start alarm, view is directory/storage occupies more than 60%
To see the specific directory situation, the main occupation of the directory has four, respectively:
/yarn 11G
/var 9.6G
/opt 4.7G
/usr 1.8G
which
Yarn mainly holds intermediate data related to Mr operation
var primarily stores log data for running and monitoring (/var/lib/cloudera-service-monitor)
Opt primarily stores CDH installation packages, including Repo,cache,parcels (where Repo,cache and run are irrelevant)
USR primarily stores installation files
Processing:
1. Adjust the monitoring log data storage path to a larger quota directory
2. Adjust yarn storage path to a larger quota directory
Hadoop Primary Node Storage alarms