Kylin relies on HBase not only for Cube, but also for meta data. A stable HBase is the basis if you expect Kylin with high performance and high availability. I don't know how EMR can ensure HBase HA when auto-scaling; if it couldn't, that would be a problem.
2017-08-09 0:19 GMT+08:00 bk_180 <gilberto.k...@gmail.com>: > Hi Shaofeng Shi, > > > Looking for some information on Hbase UI I realized that the node informed > in the log was stopped during the process. So this explaind the timeout, > correct? > > {"url":"http://<masternode>:7070/kylin/api/jobs/349d62e3- > 7748-46ee-8009-3085726b5933","exception":"java.lang.RuntimeException: > org.apache.kylin.job.exception.PersistentException: > java.net.SocketTimeoutException: callTimeout=60000, callDuration=63012: > row > '/execute/349d62e3-7748-46ee-8009-3085726b5933' on table 'kylin_metadata' > at > region=kylin_metadata,/dict/VIEW_CREDITO/COD_PROMOTOR/ > a61c83e7-01e7-4971-9539-dd5c7ed07dff.dict,1500376865680. > 301b695304237ff0d1a7774bbff83b67., > /hostname=<node>,/16020,1501960877256, seqNum=163046"} > > We are running a Cluster on AWS EMR (hbase in S3) and using the tasknodes > with autoscalling feature turned on, I´m understandig that this node was > decomissioned due this configuration. > > Can you see any other problem we migth face? > > Thanks for answering my post. > > -- > View this message in context: http://apache-kylin.74782.x6. > nabble.com/Timeout-when-refreshing-a-cube-tp8658p8671.html > Sent from the Apache Kylin mailing list archive at Nabble.com. > -- Best regards, Shaofeng Shi 史少锋