> On Nov. 10, 2016, 6:01 a.m., Amareshwari Sriramadasu wrote: > > Overall changes look fine. But I'm just wondering it would be good if we > > can have a configuration to enable/disable the feature fully - > > lens.cube.metastore.enable.datacompleteness.checks - if it is set to false, > > it can fall back to the old code, and not otherwise.
Agreed. It makes sense to add this to enable/disable the feature from this config. - Narayan ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/52514/#review155550 ----------------------------------------------------------- On Nov. 9, 2016, 1:39 p.m., Narayan Periwal wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/52514/ > ----------------------------------------------------------- > > (Updated Nov. 9, 2016, 1:39 p.m.) > > > Review request for lens. > > > Bugs: LENS-1333 > https://issues.apache.org/jira/browse/LENS-1333 > > > Repository: lens > > > Description > ------- > > Though lens has partition registration being done whenever data is available, > there is no guarantee the partition registered is complete. There can be > different ways to know if the data is complete for partition. One option > could be to have a partition property saying whether it is complete or not. > Other could be to do a http call to another hosted service and more. > > Proposal here is to add an interface for DataCompletenessChecker and do the > check while resolving partitions. > > Here are some of the capabilities we would like to add in Lens : > # Lens will check partition existence first, if it exists, then check the > completeness percentage. If the completeness percentage is less than a > configured threshold (default should be 98, 99 or even 100), Lens will fail > the query. > # Lens's accept query on partial data will accept on incomplete data as well. > # Lens will also option to override the completeness percentage threshold > value at query level > # Lens will still have look ahead capability of daily being incomplete, then > it will union with hourly. > # If same measure is there in two different facts , Lens will we pick the one > with higher availability. > # In case of completeness percentage threshold missed, Lens will respond back > with available percentage. > > > Diffs > ----- > > lens-cube/src/main/java/org/apache/lens/cube/metadata/CubeFactTable.java > fb958c3 > > lens-cube/src/main/java/org/apache/lens/cube/metadata/CubeMetastoreClient.java > e14c43f > > lens-cube/src/main/java/org/apache/lens/cube/metadata/MetastoreConstants.java > 4585ef7 > lens-cube/src/main/java/org/apache/lens/cube/parse/CandidateFact.java > 5dc9dc9 > > lens-cube/src/main/java/org/apache/lens/cube/parse/CandidateTablePruneCause.java > 78fb21d > lens-cube/src/main/java/org/apache/lens/cube/parse/CubeQueryConfUtil.java > f20f105 > lens-cube/src/main/java/org/apache/lens/cube/parse/DefaultChecker.java > PRE-CREATION > > lens-cube/src/main/java/org/apache/lens/cube/parse/MaxCoveringFactResolver.java > 13f1aa4 > > lens-cube/src/main/java/org/apache/lens/cube/parse/StorageTableResolver.java > db26718 > lens-cube/src/main/resources/olap-query-conf.xml 0c888ca > lens-cube/src/test/java/org/apache/lens/cube/parse/CubeTestSetup.java > 0c43cb5 > > lens-cube/src/test/java/org/apache/lens/cube/parse/FieldsCannotBeQueriedTogetherTest.java > 7afa32e > > lens-cube/src/test/java/org/apache/lens/cube/parse/MockCompletenessChecker.java > PRE-CREATION > lens-cube/src/test/java/org/apache/lens/cube/parse/TestBaseCubeQueries.java > f6cec1b > lens-cube/src/test/java/org/apache/lens/cube/parse/TestCubeRewriter.java > b90d4d3 > lens-cube/src/test/java/org/apache/lens/cube/parse/TestQueryRewrite.java > 0aa31f4 > lens-server/src/main/resources/lenssession-default.xml 9a8f9da > src/site/apt/admin/session-config.apt d480f88 > src/site/apt/user/olap-query-conf.apt ce35ec4 > > Diff: https://reviews.apache.org/r/52514/diff/ > > > Testing > ------- > > > Thanks, > > Narayan Periwal > >
