Details
-
Improvement
-
Status: Closed
-
Minor
-
Resolution: Duplicate
-
0.2.0
-
None
-
None
-
None
-
N/A
Description
Should be able to specify that a file be "fully" or "widely" replicated, rather than an explicit replication count. This would be useful for job configuration and jar files, and probably other files whose use is wide enough to necessitate reducing latency to access them.
The current implementation will also complain if you specify replication that is wider than the system's maximum replication value, and has no facility to enable "full" replication should the number of datanodes exceed the current maximum settable value of 32k.
Attachments
Issue Links
- is duplicated by
-
HADOOP-171 need standard API to set dfs replication = high
- Closed
- is related to
-
HADOOP-170 setReplication and related bug fixes
- Closed