[HADOOP-1627] DFSAdmin incorrectly reports cluster data.

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 0.13.0
• Fix Version/s: 0.19.0

Environment

Various minor improvements to output of 'dfsadmin -report'

• Hadoop Flags: Reviewed
• Assignee: Raghu Angadi
• Reporter: Konstantin Shvachko
• Created: 18/Jul/07 03:16
• Updated: 02/May/13 02:29
• Resolved: 14/Jul/08 18:21

Description

This is what dfsadmin -report currently prints.

hadoop:504>bin/hadoop dfsadmin -report
Total raw bytes: 1705541296128 (1.55 TB)
Used raw bytes: 1388220184240 (1.26 TB)
% used: 81.39%

Total effective bytes: 8388847 (8.00 MB)
Effective replication multiplier: 165484.0270945459
-------------------------------------------------
Datanodes available: 3

Name: alive.node.1:50077
State          : In Service
Total raw bytes: 60003868672 (55.88 GB)
Used raw bytes: 21500184005 (20.02 GB)
% used: 35.83%
Last contact: Tue Jul 17 19:58:00 PDT 2007


Name: alive.node.2:50077
State          : In Service
Total raw bytes: 1645537427456 (1.49 TB)
Used raw bytes: 1366720000235 (1.24 TB)
% used: 83.05%
Last contact: Tue Jul 17 19:58:01 PDT 2007


Name: dead.node:50077
State          : In Service
Total raw bytes: 0 (0.0 k)
Used raw bytes: 0 (0.0 k)
% used: NaN%
Last contact: Wed Nov 01 12:18:10 PST 2006

1. "Effective replication multiplier" is confusing and prints a strange number
2. Datanodes available: 3 - should count only live nodes.
3. A dead node "State" should not be "In Service".
4. "% used:" should not print "NaN%"
5. We should synchronize dfsadmin reporting with the what we report via the web UI.
   Field names and numbers should be the same.

Issue Links

• incorporates: HADOOP-632 dfs -report lists "missing" datanodes
• incorporates: HADOOP-942 dfsadmin -report returns deadnode as "In Service"
• is depended upon by: HADOOP-942 dfsadmin -report returns deadnode as "In Service"