AnsweredAssumed Answered

Issues related to Migration from MapR 4.0.1 to 4.0.2

Question asked by knaudiyal on May 19, 2015
We migrated a 5 node cluster from 4.0.1 to 4.0.2 using automated scripts provided with the installation. After completing the installation, we found couple of issues. Need guidance in how these issues could be handled.

 1. It seems that the installation scripts doesn't update the central configuration on MapRFS, due to which old version configs are being replicated into nodes on every refresh. What needs to be done, so that central config gets updated after the migration, instead of manually updating it?

 2. Hadoop Version, even after updating central configuration in maprfs://, is not available to configure.sh. When we are running configure.sh -R after the completion of migration, it creates a directory 'hadoop-' in /opt/mapr/hadoop. Configure.sh is not getting hadoop version, though correct hadoop version is being refreshed via central configuration to all the nodes. We validated hadoop_version file before running configure.sh.
 - Configure.sh seems to update  symlinks.

 4.  There are leftover symlinks in the directories pointing to 4.0.1 jars which we deleted manually.

5. Yarn-site entries were to be modified manually to point to hadoop.2.5.1 jars.

Hopefully will get some guidance in resolving this before we start our next migration.

Thanks
Kapil

Outcomes