Does the job engine instance get cube status updated automatically? If true, that indicates the configuration of “kylin.rest.servers” may still have problem, please ensure the hostname and port are properly configured; Besides, is there any connection error in Kylin log?
On 5/6/15, 3:45 PM, "Tao Wong" <[email protected]> wrote: >Hi,shaofeng >Another promble arise. > >As i did above,we have three instances, one instance to run all (job >engine >+ query engine),the other two to run query >engine only. > >However,the two instances can not get the latest information about the >cube,which means these two query instances can not to sync up with the >main instance. > >So,my problem is how to configure and control the time to sync all the >"QUERY" mode kylin instances up with the "ALL" mode kylin instance after >the cube is built succussfully? > >thanks. > >2015-05-05 17:04 GMT+08:00 Tao Wong <[email protected]>: > >> It works. >> Thank you. >> >> 2015-05-05 16:50 GMT+08:00 Shi, Shaofeng <[email protected]>: >> >>> The kylin.rest.servers need be a full list of the cluster, include >>>itself; >>> in your case all instances should have the same: >>> >>> >>>kylin.rest.servers=machinename:port1,machinename:port2,machinename:port3 >>> >>> >>> On 5/5/15, 4:47 PM, "Tao Wong" <[email protected]> wrote: >>> >>> >Hi,shaofeng >>> > >>> > >>> > >>> >*For both option 1 and 2, remember to update the kylin.rest.servers to >>> >letthem aware each other, so that they can notify the other on a cube >>> >statechange;* >>> > >>> >With option 2. but 3 instances. >>> >*kylin instance 1:* configure * kylin.rest.servers *in >>>kylin.properties >>> >with* >>> >kylin.rest.servers=machinename:port2,machinename:port3* >>> >*kylin instance 2: *configure * kylin.rest.servers *in >>>kylin.properties >>> >with *kylin.rest.servers=machinename:port1,machinename:port3* >>> >*kylin instance 3: *configure kylin.rest.servers in kylin.properties >>> with >>> >*kylin.rest.servers=machinename:port1,machinename:port2* >>> > >>> >do you mean this ? >>> > >>> >2015-05-05 16:15 GMT+08:00 Shi, Shaofeng <[email protected]>: >>> > >>> >> Dong, you might misunderstand the ³offline² here; The doc ³Off >>>Hadoop >>> >>cli >>> >> installation² talks about how to install Kylin in a machine which is >>> out >>> >> of Hadoop cluster; In that case, you need configure a remote cli >>> >> host/user/pwd in kylin.properties so that Kylin can ssh to that cli >>>to >>> >> execute hadoop/hive/hbase commands; As usually it is not a problem >>>for >>> >> user to install Kylin in hadoop cluster, the ³off cli² mode will >>>not be >>> >> our focus anymore; >>> >> >>> >> For your case, if you want to install 2 Kylin instances, there are >>>two >>> >> options in my mind: >>> >> >>> >> Option 1: >>> >> Install Kylin in two machines, and configure one instance to run >>>query >>> >> engine only, and the other run all (job engine + query engine); Yang >>> has >>> >> mentioned how to configure that in a previous email; >>> >> >>> >> Option 2: >>> >> Install the 2 instances in 1 machine but assign different ports to >>>them >>> >> (modify in tomcat/conf/server.xml); Please also do the same >>> >>configuration >>> >> as option 1; After that, use different ports to access different >>> >>instances; >>> >> >>> >> For both option 1 and 2, remember to update the kylin.rest.servers >>>to >>> >>let >>> >> them aware each other, so that they can notify the other on a cube >>> state >>> >> change; >>> >> >>> >> Please let us know if you have further question. >>> >> >>> >> On 5/5/15, 3:58 PM, "dong wang" <[email protected]> wrote: >>> >> >>> >> >shaofeng, then, does it mean that we couldn't start 2 different >>>kylin >>> >> >instances either on the same one machine or on 2 different >>>machines to >>> >> >POINT to the same CDH hadoop cluster? >>> >> > >>> >> >2015-05-05 15:51 GMT+08:00 Shi, Shaofeng <[email protected]>: >>> >> > >>> >> >> >>> >> >> examples/test_case_data/sandbox/kylin.properties, as the folder >>>name >>> >> >> indicates, it is for unit test (pointing to a HDP sandbox); >>> >> >> >>> >> >> conf/kylin.properties is for packaging a binary release; >>> >> >> >>> >> >> The offline mode is not recommended since 0.7, although it may >>>still >>> >> >>work; >>> >> >> >>> >> >> On 5/5/15, 3:44 PM, "dong wang" <[email protected]> wrote: >>> >> >> >>> >> >> >according to this email title, in which cases do the 2 >>>different >>> >> >> >kylin.properties files work? >>> >> >> > >>> >> >> >in addition, I also found that >>> >> >> > >>> >> >> >>> >> >> >>> >> >>> >> >>> >>>https://github.com/KylinOLAP/Kylin/blob/master/docs/Installation/Off%20H >>>a >>> >> >>d >>> >> >> >oop%20CLI%20Installation.md >>> >> >> >mentions some stuff about offline installation on the topic~ >>> >> >> >>> >> >> >>> >> >>> >> >>> >>> >>
