Created on 11-14-2017 12:41 PM - edited 09-16-2022 05:31 AM
Hi,
My Cluster has 3 nodes which occupies almost 870 GB.
[hdfs@XXXX bin]$ hadoop fs -du -s -h /user/hdfs
435.3 G 870.7 G /user/hdfs
However the space held by dfs.data.dir directory on single node is more than total space occupied by the cluster.
[hdfs@XXXX bigdata]$ du -sh dfs 464G dfs [hdfs@YYYY bin]$ du -sh /bigdata/dfs 746G /bigdata/dfs [hdfs@ZZZZ ~]$ du -sh /bigdata/dfs 257G /bigdata/dfs
Isn't hdfs files are stored on dfs.data.dir directory so how the space occupied is more than that of cluster?
Please help me reduce the space of dfs directory as it is in critical state.
[hdfs@XXXX finalized]$ ls -lart total 2008 drwxr-xr-x 244 hdfs hdfs 12288 Mar 7 2017 subdir0 drwxr-xr-x 258 hdfs hdfs 12288 Mar 7 2017 subdir1 drwxr-xr-x 258 hdfs hdfs 12288 Mar 10 2017 subdir2 drwxr-xr-x 258 hdfs hdfs 12288 Mar 12 2017 subdir3 drwxr-xr-x 258 hdfs hdfs 12288 Mar 14 2017 subdir4 drwxr-xr-x 258 hdfs hdfs 12288 Mar 15 2017 subdir5 drwxr-xr-x 258 hdfs hdfs 12288 Mar 15 2017 subdir6 drwxr-xr-x 258 hdfs hdfs 12288 Mar 16 2017 subdir7 drwxr-xr-x 258 hdfs hdfs 12288 Mar 17 2017 subdir8 drwxr-xr-x 258 hdfs hdfs 12288 Mar 17 2017 subdir9 drwxr-xr-x 258 hdfs hdfs 12288 Mar 17 2017 subdir10 . . .
drwxr-xr-x 258 hdfs hadoop 12288 Nov 14 03:16 subdir119
drwxr-xr-x 258 hdfs hadoop 12288 Nov 14 05:04 subdir122
drwxr-xr-x 5 hdfs hadoop 4096 Nov 14 14:25 subdir181
[hdfs@XXXX finalized]$ pwd /bigdata/dfs/dn/current/BP-939287337-10.0.0.4-1484085163925/current/finalized [hdfs@XXXX finalized]$ These subdir have more subdirs under it and finally blocks. Can I delete the older ones?
Thanks,
Shilpa
Created 11-23-2017 10:50 AM
I copied the older subdir* from dfs/dn/current/BP<...>/current/finalized to another storage mount/drive. This has not affected my data on hdfs.
Please let me know if someone thinks there is a better way than this workaround.
Thanks,
Shilpa
Created 11-23-2017 10:50 AM
I copied the older subdir* from dfs/dn/current/BP<...>/current/finalized to another storage mount/drive. This has not affected my data on hdfs.
Please let me know if someone thinks there is a better way than this workaround.
Thanks,
Shilpa