2011/4/30 Panayotis Antonopoulos <[email protected]> > > You are right the -c flag doesn't work on CDH3U0 version. > > -Dhbase.cluster.distributed=true solved the problem for the bulk upload. > > However I am having the same NPE while trying to use the TableInputFormat. > > Does anyone know what is going on? >
Make sure that, when you pass a config into TableInputFormat, that you use HBaseConfiguration.create() rather than new Configuration(). -Todd > > Regards, > Panagiotis. > > > From: [email protected] > > To: [email protected] > > Date: Mon, 25 Apr 2011 09:50:17 -0600 > > Subject: RE: Help with NPE during bulk load (completebulkload) > > > > > > Julio, > > > > We are running CDH3U0 ( hbase 0.90.1 ) so there may be some difference > if you are running 0.90.2. Running your workaround on CDH3U0 I get the > following: > > > > hadoop jar hbase-0.90.1-cdh3u0.jar completebulkload -c hbase-site.xml > input table > > usage: completebulkload /path/to/hfileoutputformat-output tablename > > > > Note the above does not run and just prints out usage information. > > > > The following is a workaround that is currently working for us: > > > > hadoop jar hbase-0.90.1-cdh3u0.jar completebulkload > -Dhbase.cluster.distributed=true input table > > > > Understood our workaround doesn't load the full hbase-site.xml but it > seems to have been a sufficient workaround for our needs. > > > > So not sure if that's a difference between the 0.90.2 and CDH3U0, but > thought it was worth mentioning. > > > > -----Original Message----- > > From: Julio Lopez [mailto:[email protected]] > > Sent: Monday, April 25, 2011 12:27 AM > > To: [email protected] > > Subject: Re: Help with NPE during bulk load (completebulkload) > > > > Andy, > > > > What are the symptoms? > > > > You also need to include in your classpath the directory where the > zookeeper config file (zoo.cfg) is located. > > > > Yes, HBASE-3714 addresses the issue discussed here. Although, it does > not fully address the NPE in > org.apache.hadoop.hbase.zookeeper.ZKConfig.parseZooCfg(ZKConfig.java:167) > for the cases when either (a) no hbase-site.xml has been loaded, or (b) an > incomplete configuration is provided, which does not specify a value for the > hbase.cluster.distributed option. > > > > Cheers, > > > > -Julio > > > > > > Begin forwarded message: > > > > > Date: April 24, 2011 10:26:38 PM EDT > > > Subject: RE: Help with NPE during bulk load (completebulkload) > > > Source: [email protected] Archives > > > Author: Andy Sautins > > > > > > Julio, > > > > > > I agree I would love to see your workaround documented. I quickly > tried adding a '-c configfile' > > > flag to our completebulkload calls and wasn't able to get it to work. > > > > > > -----Original Message----- > > > > > > Sent: Sunday, April 24, 2011 5:39 PM > > > To: [email protected] > > > Subject: Re: Help with NPE during bulk load (completebulkload) > > > > > > On Sun, Apr 24, 2011 at 4:29 PM, Andy Sautins < > [email protected]> wrote: > > >> If I'm reading it right it looks like JIRA-3714 addresses this issue. > > >> > > > Yeah, it looks like https://issues.apache.org/jira/browse/HBASE-3714 > > > > > >> Thanks for including your workaround ( adding -c flag ). That seems > to be a better > > > workaround than the one we were using. > > >> > > > > > > We should add this to the documentation Julio? Mind providing example > and where it should > > > go in the doc (I'll do the rest). > > > > > > Thanks, > > > St.Ack > > > Read more... > > > > > > > -- Todd Lipcon Software Engineer, Cloudera
