The pit that big data stepped on
When building a cluster, the data mount point of hdfs must be specified to a separate directory, not under / home, because it is likely to cause a mount point to be full during use. In that case, / home may be full, affecting the normal use of the linux system. If the storage space of multiple mount points varies greatly, it is best to set the policy according to the space instead of using the default rotation training method, or it is easy for the disk with small space to fill up first. When using impala, you should be careful not to specify the path to the root directory when creating the table, otherwise the following directory may be deleted when the table is deleted.
