Re: Unix script for identifying current active namenode in a HA cluster
hi Nitin, Thanks for the vital input around Hadoop Home addition. At times such things totally go off the radar when you have customized your own environment. As suggested I have shared this on github : https://github.com/devopam/hadoopHA apologies if there is any problem on github as I have limited familiarity with it :( regards Devopam On Wed, Nov 5, 2014 at 12:31 PM, Nitin Pawar nitinpawar...@gmail.com wrote: +1 If you can optionally add hadoop home directory in the script and use that in path, it can be used out of the box. Also can you share this on github On Wed, Nov 5, 2014 at 10:02 AM, Devopam Mittra devo...@gmail.com wrote: hi All, Please find attached a simple shell script to dynamically determine the active namenode in the HA Cluster and subsequently run the Hive job / query via Talend OS generated workflows. It was tried successfully on a HDP2.1 cluster with 2 nn, 7 dn running on CentOS 6.5. Each ETL job invokes this script first in our framework to derive the NN FQDN and then run the hive jobs subsequently to avoid failures. Takes a max. of 2 secs to execute (small cost in our case, as compared to dealing with a failure and then recalculating the NN to resubmit the job). Sharing it with you in case you can leverage the same without spending effort to code it. Do share your feedback/ fixes if you spot any. -- Devopam Mittra Life and Relations are not binary -- Nitin Pawar -- Devopam Mittra Life and Relations are not binary findActiveNameNode.sh Description: Bourne shell script
Re: Unix script for identifying current active namenode in a HA cluster
looks good to me thanks for the share On Wed, Nov 5, 2014 at 5:15 PM, Devopam Mittra devo...@gmail.com wrote: hi Nitin, Thanks for the vital input around Hadoop Home addition. At times such things totally go off the radar when you have customized your own environment. As suggested I have shared this on github : https://github.com/devopam/hadoopHA apologies if there is any problem on github as I have limited familiarity with it :( regards Devopam On Wed, Nov 5, 2014 at 12:31 PM, Nitin Pawar nitinpawar...@gmail.com wrote: +1 If you can optionally add hadoop home directory in the script and use that in path, it can be used out of the box. Also can you share this on github On Wed, Nov 5, 2014 at 10:02 AM, Devopam Mittra devo...@gmail.com wrote: hi All, Please find attached a simple shell script to dynamically determine the active namenode in the HA Cluster and subsequently run the Hive job / query via Talend OS generated workflows. It was tried successfully on a HDP2.1 cluster with 2 nn, 7 dn running on CentOS 6.5. Each ETL job invokes this script first in our framework to derive the NN FQDN and then run the hive jobs subsequently to avoid failures. Takes a max. of 2 secs to execute (small cost in our case, as compared to dealing with a failure and then recalculating the NN to resubmit the job). Sharing it with you in case you can leverage the same without spending effort to code it. Do share your feedback/ fixes if you spot any. -- Devopam Mittra Life and Relations are not binary -- Nitin Pawar -- Devopam Mittra Life and Relations are not binary -- Nitin Pawar
Re: Unix script for identifying current active namenode in a HA cluster
+1 If you can optionally add hadoop home directory in the script and use that in path, it can be used out of the box. Also can you share this on github On Wed, Nov 5, 2014 at 10:02 AM, Devopam Mittra devo...@gmail.com wrote: hi All, Please find attached a simple shell script to dynamically determine the active namenode in the HA Cluster and subsequently run the Hive job / query via Talend OS generated workflows. It was tried successfully on a HDP2.1 cluster with 2 nn, 7 dn running on CentOS 6.5. Each ETL job invokes this script first in our framework to derive the NN FQDN and then run the hive jobs subsequently to avoid failures. Takes a max. of 2 secs to execute (small cost in our case, as compared to dealing with a failure and then recalculating the NN to resubmit the job). Sharing it with you in case you can leverage the same without spending effort to code it. Do share your feedback/ fixes if you spot any. -- Devopam Mittra Life and Relations are not binary -- Nitin Pawar
Re: Unix script for identifying current active namenode in a HA cluster
Good work Devopam Mittra. *RegardsMuthupandi.K* Think before you print. On Wed, Nov 5, 2014 at 12:31 PM, Nitin Pawar nitinpawar...@gmail.com wrote: +1 If you can optionally add hadoop home directory in the script and use that in path, it can be used out of the box. Also can you share this on github On Wed, Nov 5, 2014 at 10:02 AM, Devopam Mittra devo...@gmail.com wrote: hi All, Please find attached a simple shell script to dynamically determine the active namenode in the HA Cluster and subsequently run the Hive job / query via Talend OS generated workflows. It was tried successfully on a HDP2.1 cluster with 2 nn, 7 dn running on CentOS 6.5. Each ETL job invokes this script first in our framework to derive the NN FQDN and then run the hive jobs subsequently to avoid failures. Takes a max. of 2 secs to execute (small cost in our case, as compared to dealing with a failure and then recalculating the NN to resubmit the job). Sharing it with you in case you can leverage the same without spending effort to code it. Do share your feedback/ fixes if you spot any. -- Devopam Mittra Life and Relations are not binary -- Nitin Pawar