And why kylin stores meta data in two different places at the same time ? lec ssmi <[email protected]> 于2020年11月15日周日 下午1:29写道：
> Hi: > Recently, in the process of transforming kylin, I found a problem. > My kylin's version is 2.6.2, the building engine used is spark . When > building the dimensional dictionary, kylin server will start a shell > process separately to start the spark program. > After spark creates the dictionary for each field, it stores the > dictionary information on hdfs, and then updates the cubename.json file > under the /cube path on hdfs. But in the process of creating HTable later, > the information of tablename.json will be read again, and then updated. The > problem is that the creation of Htable is in the process of kylin server, > so the ResourceStore instance is different from the spark program.In the > process of creating a dictionary, spark only updates the information on > hdfs, but does not update metadata in other places, such as mysql. The > ResourceStore instance in kylin server is in the form of JDBC, resulting in > the contents of the jdbc json file and the file on hdfs are different.Will > this be a problem? > > Thanks. >
