[ https://issues.apache.org/jira/browse/HAWQ-1548?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16250961#comment-16250961 ]
Yi Jin commented on HAWQ-1548: ------------------------------ Per my design, the first log is to output dynamic total YARN cluster capacity change, not per node value, that log is output only when total YARN capacity changes. The second log is output only when available YARN capacity for HAWQ changes, not for per node. So basically, we cannot say that is for per node. > Ambiguous message while logging hawq utilization > ------------------------------------------------ > > Key: HAWQ-1548 > URL: https://issues.apache.org/jira/browse/HAWQ-1548 > Project: Apache HAWQ > Issue Type: Improvement > Components: libyarn > Reporter: Shubham Sharma > Assignee: Lin Wen > > While YARN mode is enabled, resource broker logs two things - > - YARN cluster total resource > - HAWQ's total resource per node. > Following messages are logged > {code} > 2017-11-11 23:21:40.944904 > UTC,,,p549330,th900077856,,,,0,con4,,seg-10000,,,,,"LOG","00000","Resource > manager YARN resource broker counted YARN cluster having total resource > (1376256 MB, 168.000000 CORE).",,,,,,,0,,"resourcebroker_LIBYARN.c",776, > 2017-11-11 23:21:40.944921 > UTC,,,p549330,th900077856,,,,0,con4,,seg-10000,,,,,"LOG","00000","Resource > manager YARN resource broker counted HAWQ cluster now having (98304 MB, > 12.000000 CORE) in a YARN cluster of total resource (1376256 MB, 168.000000 > CORE).",,,,,,,0,,"resourcebroker_LIBYARN.c",785, > {code} > The second message shown above is ambiguous, After reading the sentence below > it looks like that complete Hawq cluster in whole has only 98304 MB and 12 > cores. However according to the configuration it should be 98304 MB and 12 > cores per segment server. > {code} > Resource manager YARN resource broker counted HAWQ cluster now having (98304 > MB, 12.000000 CORE) in a YARN cluster of total resource (1376256 MB, > 168.000000 CORE). > {code} > Either the wrong variables are printed or we can correct the message to > represent that the resources logged are per node. As this can confuse the > user into thinking that hawq cluster does not have enough resources. -- This message was sent by Atlassian JIRA (v6.4.14#64029)