You are viewing a plain text version of this content. The canonical link for it is here.
Posted to hdfs-dev@hadoop.apache.org by "Ewan Higgs (JIRA)" <ji...@apache.org> on 2016/12/15 09:23:58 UTC

[jira] [Resolved] (HDFS-11245) HDFS ignores HADOOP_CONF_DIR

     [ https://issues.apache.org/jira/browse/HDFS-11245?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Ewan Higgs resolved HDFS-11245.
-------------------------------
    Resolution: Invalid

HADOOP-9902 had a massive rewrite of scripts which improved the consistency and rationality of all the environment variables. These improvements have changed a few things which end up meaning users/admins need to alter their environment when switching from 2.x to 3.x.

The issue here was setting {{HADOOP_USER_CLASSPATH_FIRST}} as well as {{HADOOP_CLASSPATH}} when these are no longer needed. Also, getting the tools in the classpath is also done through {{HADOOP_OPTIONAL_TOOLS}} in {{hadoop-env.sh}} instead of {{HADOOP_CLASSPATH}} hacking.

Unsetting these variables seems to fix everything.

Possible follow on: more documentation distilling the changes required when moving from 2.x to 3.x. Especially in the context of a rolling upgrade.

> HDFS ignores HADOOP_CONF_DIR
> ----------------------------
>
>                 Key: HDFS-11245
>                 URL: https://issues.apache.org/jira/browse/HDFS-11245
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: hdfs
>    Affects Versions: 3.0.0-alpha1
>         Environment: Linux
>            Reporter: Ewan Higgs
>
> It seems that HDFS on trunk is ignoring {{HADOOP_CONF_DIR}}. On {{branch-2}} I could export {{HADOOP_CONF_DIR}} and use that to store my {{hdfs-site.xml}} and {{log4j.properties}}. But on trunk it appears to ignore the environment variable.
> Also, even if hdfs can find the {{log4j.properties}}, it doesn't seem interested in opening and loading it.
> On Ubuntu 16.10:
> {code}
> $ source env.sh
> $ cat env.sh 
> #!/bin/bash
> export JAVA_HOME=/usr/lib/jvm/java-8-oracle
> export HADOOP_HOME="$HOME"/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT
> export HADOOP_LOG_DIR="$(pwd)/log"
> PATH="$HADOOP_HOME"/bin:$PATH
> export HADOOP_CLASSPATH=$(hadoop classpath):"$HADOOP_HOME"/share/hadoop/tools/lib/*
> export HADOOP_USER_CLASSPATH_FIRST=true
> {code}
> Then I set the HADOOP_CONF_DIR:
> {code}
> $ export HADOOP_CONF_DIR="$(pwd)/conf/nn"
> $ ls $HADOOP_CONF_DIR
> hadoop-env.sh  hdfs-site.xml  log4j.properties
> {code}
> Now, we try to run a namenode:
> {code}
> $ hdfs namenode
> 2016-12-14 14:04:51,193 ERROR [main] namenode.NameNode: Failed to start namenode.
> java.lang.IllegalArgumentException: Invalid URI for NameNode address (check fs.defaultFS): file:/// has no authority.
>         at org.apache.hadoop.hdfs.DFSUtilClient.getNNAddress(DFSUtilClient.java:648)
>         at org.apache.hadoop.hdfs.DFSUtilClient.getNNAddressCheckLogical(DFSUtilClient.java:677)
>         at org.apache.hadoop.hdfs.DFSUtilClient.getNNAddress(DFSUtilClient.java:639)
>         at org.apache.hadoop.hdfs.server.namenode.NameNode.getRpcServerAddress(NameNode.java:556)
>         at org.apache.hadoop.hdfs.server.namenode.NameNode.loginAsNameNodeUser(NameNode.java:687)
>         at org.apache.hadoop.hdfs.server.namenode.NameNode.initialize(NameNode.java:707)
>         at org.apache.hadoop.hdfs.server.namenode.NameNode.<init>(NameNode.java:937)
>         at org.apache.hadoop.hdfs.server.namenode.NameNode.<init>(NameNode.java:916)
>         at org.apache.hadoop.hdfs.server.namenode.NameNode.createNameNode(NameNode.java:1633)
>         at org.apache.hadoop.hdfs.server.namenode.NameNode.main(NameNode.java:1701)
> {code}
> This is weird. We have the {{fs.defaultFS}} set:
> {code}
> $ grep -n2 fs.defaultFS $HADOOP_CONF_DIR/hdfs-site.xml
> 3-<configuration>
> 4-    <property>
> 5:        <name>fs.defaultFS</name>
> 6-        <value>hdfs://localhost:60010</value>
> 7-    </property>
> {code}
> So if isn't finding this config. Where is is looking and finding {{file:///}}?
> {code}
> $ strace -f -eopen,stat hdfs namenode 2>&1 | grep hdfs-site.xml
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/jdiff/hdfs-site.xml", 0x7f05eb6d21d0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/lib/hdfs-site.xml", 0x7f05eb6d21d0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/sources/hdfs-site.xml", 0x7f05eb6d21d0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/templates/hdfs-site.xml", 0x7f05eb6d21d0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/webapps/hdfs-site.xml", 0x7f05eb6d21d0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/jdiff/hdfs-site.xml", 0x7f05eb6d21d0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/lib/hdfs-site.xml", 0x7f05eb6d21d0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/sources/hdfs-site.xml", 0x7f05eb6d21d0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", {st_mode=S_IFREG|0664, st_size=775, ...}) = 0
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", {st_mode=S_IFREG|0664, st_size=775, ...}) = 0
> [pid 16271] open("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", O_RDONLY) = 218
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/jdiff/hdfs-site.xml", 0x7f05eb6d17e0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/lib/hdfs-site.xml", 0x7f05eb6d17e0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/sources/hdfs-site.xml", 0x7f05eb6d17e0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/templates/hdfs-site.xml", 0x7f05eb6d17e0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/webapps/hdfs-site.xml", 0x7f05eb6d17e0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/jdiff/hdfs-site.xml", 0x7f05eb6d17e0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/lib/hdfs-site.xml", 0x7f05eb6d17e0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/sources/hdfs-site.xml", 0x7f05eb6d17e0) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", {st_mode=S_IFREG|0664, st_size=775, ...}) = 0
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", {st_mode=S_IFREG|0664, st_size=775, ...}) = 0
> [pid 16271] open("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", O_RDONLY) = 218
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/jdiff/hdfs-site.xml", 0x7f05eb6d2070) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/lib/hdfs-site.xml", 0x7f05eb6d2070) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/sources/hdfs-site.xml", 0x7f05eb6d2070) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/templates/hdfs-site.xml", 0x7f05eb6d2070) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/webapps/hdfs-site.xml", 0x7f05eb6d2070) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/jdiff/hdfs-site.xml", 0x7f05eb6d2070) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/lib/hdfs-site.xml", 0x7f05eb6d2070) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/sources/hdfs-site.xml", 0x7f05eb6d2070) = -1 ENOENT (No such file or directory)
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", {st_mode=S_IFREG|0664, st_size=775, ...}) = 0
> [pid 16271] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", {st_mode=S_IFREG|0664, st_size=775, ...}) = 0
> [pid 16271] open("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", O_RDONLY) = 218
> {code}
> So it's ignoring {{HADOOP_CONF_DIR}}. We can work around it using {{-conf $(pwd)/conf/nn/hdfs-site.xml}}:
> {code}
> $ strace -f -eopen,stat hdfs namenode -conf $(pwd)/conf/nn/hdfs-site.xml 2>&1 | grep hdfs-site.xml
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/jdiff/hdfs-site.xml", 0x7f9f60afb1d0) = -1 ENOENT (No such file or directory)
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/lib/hdfs-site.xml", 0x7f9f60afb1d0) = -1 ENOENT (No such file or directory)
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/sources/hdfs-site.xml", 0x7f9f60afb1d0) = -1 ENOENT (No such file or directory)
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/templates/hdfs-site.xml", 0x7f9f60afb1d0) = -1 ENOENT (No such file or directory)
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/common/webapps/hdfs-site.xml", 0x7f9f60afb1d0) = -1 ENOENT (No such file or directory)
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/jdiff/hdfs-site.xml", 0x7f9f60afb1d0) = -1 ENOENT (No such file or directory)
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/lib/hdfs-site.xml", 0x7f9f60afb1d0) = -1 ENOENT (No such file or directory)
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/sources/hdfs-site.xml", 0x7f9f60afb1d0) = -1 ENOENT (No such file or directory)
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", {st_mode=S_IFREG|0664, st_size=775, ...}) = 0
> [pid 16493] stat("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", {st_mode=S_IFREG|0664, st_size=775, ...}) = 0
> [pid 16493] open("/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/share/hadoop/hdfs/templates/hdfs-site.xml", O_RDONLY) = 218
> [pid 16493] stat("/home/ehigg90120/src/hadoop-run/tutorial/conf/nn/hdfs-site.xml", {st_mode=S_IFREG|0644, st_size=2107, ...}) = 0
> [pid 16493] open("/home/ehigg90120/src/hadoop-run/tutorial/conf/nn/hdfs-site.xml", O_RDONLY) = 218
> ------8<------
> {code}
> Great! However, it's not finding my  log4j.properties for some reason. This is annoying because hdfs isn't printing anything or logging anywhere. Where is it looking?
> {code}
> $ strace -f hdfs namenode -conf $(pwd)/conf/nn/hdfs-site.xml 2>&1 | grep log4j.properties
> stat("/home/ehigg90120/src/hadoop-run/tutorial/conf/nn/log4j.properties", {st_mode=S_IFREG|0644, st_size=13641, ...}) = 0
> {code}
> It found it, but it only statted it. It never opened it! So it seems there's at least one bug here where {{log4j.properties}} is being ignored. But shouldn't {{HADOOP_OPTS}} be set and configuring it to print to the console and to my log dir?
> {code}
> $ hdfs --debug namenode -conf $(pwd)/conf/nn/hdfs-site.xml 2>&1 | grep HADOOP_OPTS
> DEBUG: HADOOP_OPTS accepted -Dhdfs.audit.logger=INFO,NullAppender
> DEBUG: Appending HDFS_NAMENODE_OPTS onto HADOOP_OPTS
> DEBUG: HADOOP_OPTS accepted -Dyarn.log.dir=/home/ehigg90120/src/hadoop-run/tutorial/log
> DEBUG: HADOOP_OPTS accepted -Dyarn.log.file=hadoop.log
> DEBUG: HADOOP_OPTS accepted -Dyarn.home.dir=/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT
> DEBUG: HADOOP_OPTS accepted -Dyarn.root.logger=INFO,console
> DEBUG: HADOOP_OPTS accepted -Djava.library.path=/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/lib/native
> DEBUG: HADOOP_OPTS accepted -Dhadoop.log.dir=/home/ehigg90120/src/hadoop-run/tutorial/log
> DEBUG: HADOOP_OPTS accepted -Dhadoop.log.file=hadoop.log
> DEBUG: HADOOP_OPTS accepted -Dhadoop.home.dir=/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT
> DEBUG: HADOOP_OPTS accepted -Dhadoop.id.str=ehigg90120
> DEBUG: HADOOP_OPTS accepted -Dhadoop.root.logger=INFO,console
> DEBUG: HADOOP_OPTS accepted -Dhadoop.policy.file=hadoop-policy.xml
> DEBUG: HADOOP_OPTS declined -Dhadoop.security.logger=INFO,NullAppender
> DEBUG: Final HADOOP_OPTS: -Djava.net.preferIPv4Stack=true -Dhdfs.audit.logger=INFO,NullAppender -Dhadoop.security.logger=INFO,RFAS -Dyarn.log.dir=/home/ehigg90120/src/hadoop-run/tutorial/log -Dyarn.log.file=hadoop.log -Dyarn.home.dir=/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT -Dyarn.root.logger=INFO,console -Djava.library.path=/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT/lib/native -Dhadoop.log.dir=/home/ehigg90120/src/hadoop-run/tutorial/log -Dhadoop.log.file=hadoop.log -Dhadoop.home.dir=/home/ehigg90120/src/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha2-SNAPSHOT -Dhadoop.id.str=ehigg90120 -Dhadoop.root.logger=INFO,console -Dhadoop.policy.file=hadoop-policy.xml
> {code}
> So it seems it is being configured and passed to the namenode. It's just not obeying it as far as I can see. 
> So maybe there are two possibly related bugs:
> 1. {{HADOOP_CONF_DIR}} is ignored
> 2. The logger is not using {{log4j.properties}} or the command line. I would expect it to use the {{log4j.properties}} in the {{HADOOP_CONF_DIR}}.
> I feel like I must be misunderstanding something since this seems like a pretty big issue but I didn't find any open tickets about it or any tickets describing a new way of configuring clusters.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-dev-unsubscribe@hadoop.apache.org
For additional commands, e-mail: hdfs-dev-help@hadoop.apache.org