2010/02/23

Cloudera Desktop setup tip: DNS, DNS, DNS

Yes, it's important with Cloudera Desktop, just like everything else - try to fudge domain names and get bit.

Strange errors like:

"An unknown error occurred: hdfs put returned bad code: 255 stderr: 10/02/18 18:41:52 INFO ipc.Client: Retrying connect to server: localhost/127.0.0.1:8020... Bad connection to FS. command aborted."

When attempting to upload a file - and I know that localhost is indeed responding properly on that port because it works find from the command line.

Solution: Spend the few minutes to determine what the real domain names are (*1) and set up the config files (*2) properly.

*1: ifconfig will tell you the IP address of the node you're on; host will tell you the domain name. If there is none, see *3 below.

*2: config files of potential interest:
/usr/share/cloudera-desktop/conf/cloudera-desktop.ini
/etc/hadoop/conf/masters
/etc/hadoop/conf/slaves
/etc/hadoop/conf/core-site.xml
/etc/hadoop/conf/hdfs-site.xml
/etc/hadoop/conf/mapred-site.xml
/etc/hadoop/conf/hadoop-metrics.properties
/etc/hadoop/conf/hadoop-env.sh
/etc/hadoop/conf/configuration.xsl
/etc/hadoop/conf/fair-scheduler.xml

*3: No domain name? See whoever's in charge of DNS, to fix it. That you? Well, you can either do it right (a little effort up front, pays big in the long run...) or you can try to handle it via /etc/hosts - good luck with that though.

Note to self: Don't shortcut DNS!