Ted Yu created SLIDER-399:
-----------------------------
Summary: Agent command doesn't consider the case where LOG_DIRS
contains multiple directories
Key: SLIDER-399
URL: https://issues.apache.org/jira/browse/SLIDER-399
Project: Slider
Issue Type: Bug
Reporter: Ted Yu
On a cluster where LOG_DIRS contains multiple directories, I saw the following
in AM log:
{code}
14/09/04 21:43:48 INFO appmaster.RoleLaunchService: Starting container with
command: python ./infra/agent/slider-agent/agent/main.py --label
container_1407274006606_0097_01_000006___HBASE_THRIFT --zk-quorum
hor10n07.gq1.ygridcore.net:2181 --zk-reg-path
/registry/org-apache-slider/hbcluster27 > $LOG_DIRS/agent.out 2>&1 ;
{code}
The above command resulted in:
{code}
Stack trace: ExitCodeException exitCode=1: /bin/bash:
/grid/0/hdp/yarn/log/application_1407274006606_0097/container_1407274006606_0097_01_000002,/grid/1/hdp/yarn/log/application_1407274006606_0097/container_1407274006606_0097_01_000002,/grid/2/hdp/yarn/log/application_1407274006606_0097/container_1407274006606_0097_01_000002,/grid/3/hdp/yarn/log/application_1407274006606_0097/container_1407274006606_0097_01_000002,/grid/4/hdp/yarn/log/application_1407274006606_0097/container_1407274006606_0097_01_000002,/grid/5/hdp/yarn/log/application_1407274006606_0097/container_1407274006606_0097_01_000002/agent.out:
No such file or directory
{code}
The cause seems to be that $LOG_DIRS is the concatenation of 6 directories.
$LOG_DIRS/agent.out would not be accessible, leading to container launch
failure.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
