Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Won't Fix
-
1.5.2
-
None
-
None
Description
The tar ball for 1.5.2 with Hadoop 1 (the default) has the scala version appended to the file name, which leads to spark_ec2.py failing on starting up the cluster. Here is the record from the S3 contents:
<Contents>
<Key>spark-1.5.2-bin-hadoop1-scala2.11.tgz</Key>
<LastModified>2015-11-10T06:45:17.000Z</LastModified>
<ETag>"056fc68e549db27d986da707f19e39c8-4"</ETag>
<Size>234574403</Size>
<StorageClass>STANDARD</StorageClass>
</Contents>
Maybe one could provide one without the scala suffix (default)?
A workaround is to use set the Hadoop version to a version different from 1 when calling spark-ec2.