Re: query return error result.

I checked the code, and find root cause is DumpMerger.enqueueFromDump() I create a jira KYLIN-2926 <https://issues.apache.org/jira/browse/KYLIN-2926> to trace the bug. 2017-10-09 10:37 GMT+08:00 yu feng : > the cube is using hllc15, we are tracing the code and try to find the

Re: query return error result.

the cube is using hllc15, we are tracing the code and try to find the reason. 2017-10-08 14:52 GMT+08:00 Li Yang : > Interesting... is it HLL count distinct or bitmap count distinct? > > On Wed, Sep 27, 2017 at 11:19 AM, yu feng wrote: > >> I add some log and find data from

Re: query return error result.

I add some log and find data from hbase is incorrect. 2017-09-27 11:17 GMT+08:00 yu feng : > I have a cube like this : > dimensions : source_type, source_id, name, dt > measures:count(distinct uid), count(1) , count(distinct buyer) > > I run the query : > > select source_

query return error result.

I have a cube like this : dimensions : source_type, source_id, name, dt measures:count(distinct uid), count(1) , count(distinct buyer) I run the query : select source_type, source_id, name, count(distinct uid), count(uid) as cnum, count(distinct buyer) as buyerNum, count(buyer) as bnum fr

Re: different mr config for different project or cube

Great， Thanks. 2017-09-06 15:07 GMT+08:00 Shuangyin Ge : > Feng, > > Can you please take a look at this part of tutorial http://kylin.apache. > org/docs20/install/advance_settings.html? > > 2017-09-06 14:01 GMT+08:00 yu feng : > >> How to do it in GUI? by cube level

different mr config for different project or cube

I remember Kylin support use different mr config file for different project, like KYLIN-1706 and KYLIN-1706 However, I do not know how to use it in kylin-2.0.0. it would be appreciated if anyone

Kylin query return empty after upgrade to 2.0

Hi, After I upgrade our env. from 1.5.2.1 to 2.0.0, building things go well, However, every query to older(generated before upgrading) return empty(0 result), and newly build segment return result success. I add some debug log and find in GTFilterScanner, I get 0 input from inputIterator and retur

Re: How can we refresh auth token for data source?

we always start a crontab job to refresh the keberos token periodically. 2017-05-02 17:07 GMT+08:00 Kanta Kuramoto : > We use kerberized hadoop cluster for datasource. > It seems Kylin server loads auth token only when it starts, so after the > token is expired, Kylin server can not load metadat

Re: MAX aggregation for name column

we also has this demand, for derived dimension, lookup table is too big for memory, so put the name to measure is a good choice, I will work on it to support max on string. 2016-06-26 21:24 GMT+08:00 Li Yang : > The best solution is to let Kylin support max on strings. Please open a > JIRA to tra

Re: 答复: kylin.job.mr.lib.dir配置在hdfs下无效问题

I think KYLIN-1021 has done about setting a HDFS path that store all jars, BTW, I think the directory is setting to ${kylin.hdfs.working.dir}/dependencies automatically is a better idea. 在 2016年6月7日 下午4:28，Li Yang 写道： > Agree allowing HDFS path as well is a good idea. Please open a JIRA. > > 2016

Re: Does cube data of Hbase always exist in memory ?

Precalculated date of one cube stored in hbase, and kylin access them by scan operator, so they are managed by hbase I think. 2016-05-04 16:36 GMT+08:00 王猛 : > Does cube data of Hbase always exist in memory ? Or depend on Hbase > tactics ? > Thanks >

Re: Encountered "day" at line 1, column 8. Was expecting one of: "UNION" ... "INTERSECT" ..

Because day is a keyword in calcite(such as extract(day from col)),so you need use double quote on both sides of day, so in kylin all column is capital word, you need write this column like "DAY", year、month are the same format. hope it works for you. 2016-03-23 11:47 GMT+08:00 Bin.Li : > hello

Re: java.lang.IllegalArgumentException: Too many digits for NumberDictionary

I has encountered this problem before, Now you can modify the column to string type(by create a view and cast this column) to skip this problem, My suggestion is try to use TrieDictionary if there is some error while building NumberDictionary or DateStrDictionary, This will more robust. 2016-03-0

Re: NPE on step 2

you should check the log of this mapreduce job. maybe you can find some hints. 2016-03-07 17:27 GMT+08:00 热爱大发挥 <385639...@qq.com>: > nullPointException happen on step 2 > > #2 Step Name: Extract Fact Table Distinct Columns > > 2016-03-07 03:23:03.889 - State of Hadoop job: > job_1457321012016_00

Re: kylin's compression problem when I comment the related items

you should check some file like 'conf/kylin_job_conf.xml', those 'set key = value' read from this file. 2016-02-23 17:54 GMT+08:00 Mars J : > Hello, > I ran into a problem as the subject. Details as follow: > > version of my envr > OS , CentOS 7 > Apache Hadoop 2.7.2 > Apache Hive 1.2.1 > Apa

Re: about "Table snapshot size limit"

- 原始邮件 ------ > *发件人:* "yu feng"; > *发送时间:* 2016年2月15日(星期一) 中午11:10 > *收件人:* "user"; > *主题:* Re: about "Table snapshot size limit" > > kylin.table.snapshot.max_mb is the max size of files that the lookup table > location, you can increase the

Re: about "Table snapshot size limit"

kylin.table.snapshot.max_mb is the max size of files that the lookup table location, you can increase the config value and skip the error. 2016-02-15 9:42 GMT+08:00 热爱大发挥 <385639...@qq.com>: > I do not understand why limit the size of the table 300MB, I think 300MB > is too small, Can you explai

Re: Two cubes on same FACT table

On Wed, Jan 13, 2016 at 8:39 PM, yu feng wrote: > >> You have to use table name in your query, if you want to query >> on specific cube, you can create two view on the same fact table with >> different name and create cube view different view, If your query match two >> cub

Re: Two cubes on same FACT table

You have to use table name in your query, if you want to query on specific cube, you can create two view on the same fact table with different name and create cube view different view, If your query match two cube, kylin do not know which one is you want, so maybe result can not make you satisfied.

Re: Issue about Kerberos

the log stack tells "Connection refused", maybe you should check rm address config in hadoop configuration(yarn-sitre.xml), a simple MR job will be a good test. 2016-01-13 23:26 GMT+08:00 Luke Han : > Hi Sudeep, >Have you tried with none-Kerberos cluster? >Thanks. > > Luke > > > Best Rega