[ https://issues.apache.org/jira/browse/AMBARI-15273?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15435273#comment-15435273 ]
Mark S commented on AMBARI-15273: --------------------------------- Just to follow up. You can configure the $\{hadoop.tmp.dir\} property via the web interface http://localhost:8080/#/main/services/HDFS/configs, and add the entry to "Custom core-site" via the "Add Property" link. I was able to verify this property being set correctly with the following command: {code} sudo -u hdfs /usr/hdp/current/hadoop-client/bin/hadoop org.apache.hadoop.conf.Configuration | grep hadoop.tmp.dir <property><name>hadoop.tmp.dir</name><value>/data/hadoop-tmp/${user.name}</value><source>core-site.xml</source></property> <property><name>fs.s3.buffer.dir</name><value>${hadoop.tmp.dir}/s3</value><source>core-default.xml</source></property> <property><name>fs.s3a.buffer.dir</name><value>${hadoop.tmp.dir}/s3a</value><source>core-default.xml</source></property> <property><name>io.seqfile.local.dir</name><value>${hadoop.tmp.dir}/io/local</value><source>core-default.xml</source></property> {code} > Cannot set ${hadoop.tmp.dir} > ---------------------------- > > Key: AMBARI-15273 > URL: https://issues.apache.org/jira/browse/AMBARI-15273 > Project: Ambari > Issue Type: Bug > Affects Versions: 2.1.2 > Environment: CentOS Linux release 7.2.1511 (Core) > ambari-agent-2.1.2-377.x86_64 > ambari-server-2.1.2-377.x86_64 > ambari-metrics-hadoop-sink-2.1.2-377.x86_64 > ambari-metrics-monitor-2.1.2-377.x86_64 > ambari-metrics-collector-2.1.2-377.x86_64 > HDFS 2.7.1.2.3 > MapReduce2 2.7.1.2.3 > YARN 2.7.1.2.3 > Reporter: Mark S > > I am trying to set the "hadoop.tmp.dir" property, as a work-around for a > situation where my Linux "/tmp" partition is running out of space and causing > my Hadoop MapReduce jobs to fail. > Currently the following folders are causing my tmp partition to fill up and > my Hadoop Map Reduce to fail, > {code} > /tmp/hadoop-user/mapred/local > /tmp/hadoop-user/mapred/staging > {code} > and I am pretty sure is set by the following two properties: > {code} > # > https://hadoop.apache.org/docs/r2.7.1/hadoop-mapreduce-client/hadoop-mapreduce-client-core/mapred-default.xml > mapreduce.cluster.local.dir=${hadoop.tmp.dir}/mapred/local > mapreduce.jobtracker.staging.root.dir=${hadoop.tmp.dir}/mapred/staging > {code} > Thus far, I have been trying unsuccessfully to set the "hadoop.tmp.dir" > property using the following command and restarting all my services after > this. > {code} > # https://cwiki.apache.org/confluence/display/AMBARI/Modify+configurations > /var/lib/ambari-server/resources/scripts/configs.sh -u admin -p admin set > <ambari_host> <ambari_cluster_name> "core-site" "hadoop.tmp.dir" > "/data/hadoop-tmp/\${user.name}" > {code} -- This message was sent by Atlassian JIRA (v6.3.4#6332)