Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
5.1.0, 5.2.0
-
None
-
None
Description
The change in https://issues.apache.org/jira/browse/OOZIE-2791 hardcoded replication factor to be 3 when using multi-threaded copying while single-threaded copy will follow the replication factor in hdfs-site.xml.
This could be problematic when a cluster has less than 3 data nodes. Since the replication can never be 3 in this case, hdfs report will show a lot of files are under replicated.
$ hdfs dfsadmin -report | head Configured Capacity: 148067303424 (137.90 GB) Present Capacity: 147612037120 (137.47 GB) DFS Remaining: 145914187776 (135.89 GB) DFS Used: 1697849344 (1.58 GB) DFS Used%: 1.15% Under replicated blocks: 1003 Blocks with corrupt replicas: 0 Missing blocks: 0 Missing blocks (with replication factor 1): 0 Pending deletion blocks: 0
And the message from hdfs fsck will be like
/user/oozie/share/lib/lib_20200707223334/git/commons-codec-1.10.jar: Under replicated BP-1985902824-10.65.207.110-1594161186186:blk_1073742826_2002. Target Replicas is 3 but found 2 live replica(s), 0 decommissioned replica(s) and 0 decommissioning replica(s). /user/oozie/share/lib/lib_20200707223334/git/commons-lang3-3.3.2.jar: Under replicated BP-1985902824-10.65.207.110-1594161186186:blk_1073742810_1986. Target Replicas is 3 but found 2 live replica(s), 0 decommissioned replica(s) and 0 decommissioning replica(s). /user/oozie/share/lib/lib_20200707223334/git/httpclient-4.5.9.jar: Under replicated BP-1985902824-10.65.207.110-1594161186186:blk_1073742815_1991. Target Replicas is 3 but found 2 live replica(s), 0 decommissioned replica(s) and 0 decommissioning replica(s).