[ https://issues.apache.org/jira/browse/IMPALA-7098?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16498560#comment-16498560 ]
Dan Hecht edited comment on IMPALA-7098 at 6/1/18 8:59 PM: ----------------------------------------------------------- What value does mBlockSize on the return of hdfsGetPathInfo() give? Does it give the block group size? (And for parquet, the parquet scanner will find the row-groups and assign them to the scan range that contains the mid-point offset of the row-group. Ideally, what we'd want is to align scan ranges with row-groups. It would be worse to create scan ranges larger than a row-group. It doesn't really hurt to have a scan range smaller than a row-group.) was (Author: dhecht): What value does mBlockSize on the return of hdfsGetPathInfo() give? Does it give the block group size? > Re-enable blocksize-related tests under EC > ------------------------------------------ > > Key: IMPALA-7098 > URL: https://issues.apache.org/jira/browse/IMPALA-7098 > Project: IMPALA > Issue Type: Sub-task > Components: Frontend, Infrastructure > Affects Versions: Impala 3.1.0 > Reporter: Tianyi Wang > Assignee: Taras Bobrovytsky > Priority: Major > > In EC the unit of a scan range is a block group. So in our mini cluster a > scan range is of 3X the size of a regular block. This breaks some tests under > EC. One way to address this problem is to shrink the blocks to 1/3 of the > original size in those tests. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-all-unsubscr...@impala.apache.org For additional commands, e-mail: issues-all-h...@impala.apache.org