Hi Shaofeng, day_time type is String, and it is partition col. But I confirmed that 2018-03-01 have records. Problem comes when quering with the setting day_time “shard by” to true. If set all dimensions to false no problem for querying.
Maybe I’m not fully understand “shard by” function clearly. Your explanation for it(when to use and how to use “shard by”) is highly appreciated. Thanks. I will follow your suggestions and test on Monday. After that I will report my feedback. | | 赵希欣 | | 邮箱：[email protected] | 签名由 网易邮箱大师 定制 On 03/10/2018 22:31, ShaoFeng Shi wrote: Hi Xixin, I'm analyzing this, which should be a bug. As I don't have the model and table definition, could you please provide the data type of the col "day_time"? And I think "day_time" is the partition column, and in your cube, there was no record for the day '2018-03-01' when you run the query, can you confirm this? BTW, using the partition column as the "shared by" column is not practical, as the cardinality is low, which would downgrade the performance. Please unset that and using another high cardinality col for sharding. Thanks for your feedback! 2018-03-09 18:28 GMT+08:00 ShaoFeng Shi <[email protected]>: Hi Xixin, JIRA created for this issue: https://issues.apache.org/jira/browse/KYLIN-3285 Please book it and let's track there; thank you! 2018-03-09 12:46 GMT+08:00 xixin <[email protected]>: Hi Shaofeng, Please see the attachments:Kylin.log and cube.json the problem maybe from the "shard by" setting, after I set the day_time "shard by” to true,the issue comes when I query. If set all shard by to false, query can be done successfully. 在 2018年3月8日，20:48，ShaoFeng Shi <[email protected]> 写道： Hi Xixin, Thanks for reporting. Would you please share the kylin.log and how you define the Cube? If you can reproduce the error with the sample cube "kylin_sales", that would be appreciated. 2018-03-08 14:03 GMT+08:00 xixin <[email protected]>: I did a test, when I set mandatory dimension on aggression groups lead this error 在 2018年3月8日，11:54，xixin <[email protected]> 写道： Hi Kylin team, Kylin version update from 2.2.0 to 2.3.0; cube was build before update, and query without error. After updatge to version 2.3.0 and query "select count(userid) num,day_time from record_ap group by day_time LIMIT 1000” show errors below: Column 0 value '2018-03-06' met dictionary error: Value '2018-03-06' (2018-03-06) not exists! while executing SQL: "select count(userid) num,day_time from record_ap group by day_time LIMIT 1000” Where '2018-03-06’ come from? I was fully confused. Cube have several segments: Starttime-endtime 20180101-20180225 20180225-20180304 20180304-20180305 20180305-20180306 20180306-20180307 -- Best regards, Shaofeng Shi 史少锋 -- Best regards, Shaofeng Shi 史少锋 -- Best regards, Shaofeng Shi 史少锋
