[ https://issues.apache.org/jira/browse/KYLIN-3371?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16468525#comment-16468525 ]
Roberto Tardío Olmos edited comment on KYLIN-3371 at 5/9/18 8:04 AM: --------------------------------------------------------------------- We have not configured "kylin.job.scheduler.poll-interval-second" property either in Kylin 2.2 or in Kylin 2.3.1. Therefore, we use the default value. We could try to configure this property with a smaller time vale, such us 5 seconds. As a result of this, I checked in git the code of Kylin but I do not see anything strange (considering that I am a newbie in Kylin's code :D) was (Author: rober tardio): We have not configured "kylin.job.scheduler.poll-interval-second" property either in Kylin 2.2 or in Kylin 2.3.1. Therefore, we use the default value. We could try to configure to property with a smaller time vale, such us 5 seconds. As a result of this, I checked in git the code of Kylin but I do not see anything strange (considering that I am a newbie in Kylin's code :D) > Increase of time between steps after Kylin 2.3.1 update > ------------------------------------------------------- > > Key: KYLIN-3371 > URL: https://issues.apache.org/jira/browse/KYLIN-3371 > Project: Kylin > Issue Type: Bug > Components: Job Engine > Affects Versions: v2.3.1 > Environment: -EMR 5.7 (S3 storage for Hive and HBase) > -Kylin 2.3.1 (updated from 2.2). Deployed on separated EC2 instance. > Reporter: Roberto Tardío Olmos > Priority: Major > Attachments: Screenshot1.png, Screenshot2.png > > > Hi, > We have updated from Kylin 2.2 to Kylin 2.3.1. We performed the update > process with successful. However, we have experimented an increase of time > wasted between steps during building time. For example, the time between step > 7 “Build N-Dimension Cuboid:level 1” and “…level2” is now about 30 seconds, a > lot of time if we take into account that in these steps no processing is > executed since we use the "layer" algorithm (Screenshot 1). This time between > steps is much longer than when we used Kylin 2.2, were average time was about > 5 seconds. (Screenshot 2). > Could it be related to a change in the code? The configuration we use in > Kylin 2.3.1 is the same as we used in Kylin 2.2. I have considered this JIRA > as a possible BUG. > Thanks in advance, > > Screenshot 1: > !Screenshot1.png! > > Screenshot 2 > > !Screenshot2.png! > > > A sample of the generated log. > Kylin 2.2 > > 2018-05-04 14:25:40,797 INFO [Scheduler 542798956 Job > cb843ecc-5062-49ec-a881-215097e4501a-518] execution.ExecutableManager:421 : > job id:cb843ecc-5062-49ec-a881-215097e4501a from RUNNING to READY > 2018-05-04 14:25:45,496 INFO [Scheduler 542798956 Job > cb843ecc-5062-49ec-a881-215097e4501a-518] execution.AbstractExecutable:111 : > Executing AbstractExecutable (BUILD CUBE - Captacion_Resumido_Cubo_v3 - > 20180504000000_20180505000000 - GMT+02:00 2018-05-04 14:21:34) > > Kylin 2.3.1 > > 2018-05-07 15:25:26,338 INFO [Scheduler 1942482192 Job > 6d616841-1c4d-46ca-a1b9-a18047fbada5-1006] execution.ExecutableManager:411 : > job id:6d616841-1c4d-46ca-a1b9-a18047fbada5 from RUNNING to READY > 2018-05-07 {color:#FF0000}15:25:48,365{color} INFO [Scheduler 1942482192 Job > 6d616841-1c4d-46ca-a1b9-a18047fbada5-1006] execution.AbstractExecutable:147 : > Executing AbstractExecutable (BUILD CUBE - Captacion_Resumido_Cubo_v3 - > 20180507000000_20180508000000 - GMT+02:00 2018-05-07 15:19:41) -- This message was sent by Atlassian JIRA (v7.6.3#76005)