@Jeff Watson Have you tried to log the generated upsert statement and execute it manually using sqlline? As for the versions of libraries, you need to keep in mind that Phoenix in HDP has number of patches and features comparing to the Apache Phoenix 4.7 release. If you are building your application for HDP I would suggest to use the corresponding libraries (whether it's Phoenix or HBase or Hadoop) that are located at http://nexus-private.hortonworks.com/nexus/content/groups/public If you are using maven, I would suggest to add it to the repositories section in your pom.xml : <repository> <id>public</id> <name>hwx</name> <url>http://nexus-private.hortonworks.com/nexus/content/groups/public/</url> </repository> and use the corresponding versions of phoenix/hbase/hadoop libraries. For example hadoop versions is supposed to be 2.7.3.2.5.3.0-37 And two words why CSV classes are different. Phoenix is using shading for most of the external libraries, so classes may be modified when added to the final jar (to reflect shaded imports). But it should not affect the debug information. ... View more

That's definitely strange. Just tried a fresh install of 5.1.14 with sandbox 2.5 on my Windows 8 laptop and everything works just fine without problems. Have you tried to start it with other setting for network adapter (I would suggest to try "Not attached") to check whether it's really adapter settings problem or something else? ... View more

Check what you have 'NAT' adapter in the Network settings for the VM. ... View more

I'm not sure that I understand the changed you have done with hbase-site.xml. Could you please explain it a bit more. If you can provide the master log that may be useful to understand what exactly wrong on your side. One small note. From the sandbox perspective there is no difference between running it on pure Linux box or Linux VM on MacOS. It's isolated by docker container which has it's own IP which supposed to be resolved to sandbox.hortonworks.com FQDN and this name is used by all HDP services. It's not recommended to change this name nor change it to localhost until you want to do fresh install of HDP. Port forwarding is just a way to get access to this docker container from outside. ... View more

As Josh already mentioned, the best way to deploy HDP clusters is Ambari. Bigtop is a packaging project that has some ability to deploy unmanaged cluster for testing purposes. Ambari on the other hand is the cluster managing software. It may come that one day we will see Ambari as one of the packages in Bigtop, but even in this case the deployment of cluster will be done by Ambari blueprints. ... View more

If I correctly understand, you need to change :part_key to :key. ... View more

try to turn on the java kerberos logging by adding -Dsun.security.krb5.debug=true to the HADOOP_OPTS. Usually it helps to understand what exactly fails during GSS initialization. ... View more

I checked the script more carefully. It looks like adding export CLASSPATH=$CLASSPATH:/etc/hbase/conf to zeppelin-env.sh using Zeppelin configuration in Ambari would be enough. ... View more

I bet that HBase is healthy. Was able to reproduce it locally. The same exception running test from zeppelin with perfectly working sqlline (as well as hbase shell). ... View more

@Bala Vignesh N V what do you mean by 'Fixed width file' ? Can you give a more detailed example? ... View more

Could you please explain more why distcp hftp -> hdfs didn't work? This is the recommended way to copy between different versions of hadoop. And it should be running on destination cluster. ... View more

I would also add to Ted's comment that netty-3.2.4 (which is required by hadoop-hdfs) and netty-all-4.0.23 are not conflicting because 3.2.4 is the old version with org.jboss.netty package while 4.0.23 has io.netty package. Actually they even have different artifact IDs (netty vs netty-all). So it's safe to use both of them in the same project. ... View more

you need to use the same command grant 'sami', 'RWXCA' but you need to run hbase shell with hbase kerberos ticket. Ranger is centralized platform to manage security on your cluster at one place. ... View more

This is just an information message during the connection. If your application doesn't work, check that zookeeper as well as HBase are started. ... View more

check this topic how to use doAs: https://community.hortonworks.com/questions/46500/spark-cant-connect-to-hbase-using-kerberos-in-clus.html ... View more

Are you using secured cluster? ... View more

MR job creates 1 reducer per region. So if you loading data to an empty table you may presplit table from HBase shell or use salting during table creation. ... View more

Can you run ls -la /usr/hdp/2.5.0.0-1245/hbase/lib/*.jar and attach output here? ... View more

I would suggest to return back the original protobuf (there was a thread about a similar issue somewhere on the internet, but this issue is not relevant to the protobuf). And check that no new jars were added to the HBase lib directory (like phoenix-client for example). ... View more

Phoenix supports the only way to define table with schema by using dot between. Colons are using internally by HBase and Phoenix does this transformation on its own. As I mentioned earlier Phoenix has a bug that both schema and table name should have double quotes, otherwise table created incorrectly with empty schema. Moreover on the HBase layer this cause ambiguous physical table mapping like you are seeing in your example. So, my recommendation is to use quotes for schema and table name separately or just do not use quotes. And you are right, the feature was backported from 4.8. ... View more

Try String tableStr ="ACME.ENDPOINT_STATUS"; No double quotes and dot between schema name and table name. Hope that will work. One more note: if you create table using double quotes, put schema and table name in quotes. Like create table "ACME"."ENDPOINT_STATUS" in this case you will need to quote table name like you do in your sample. It seems that there is a bug in phoenix code. If you quote them together, Phoenix creates a table with empty schema and the table name as "schema.table". You may check it by querying SYSTEM.CATALOG table. ... View more

You need to be sure that phoenix client has hbase-site.xml in the classpath. You may do this by setting HBASE_CONF_DIR environment variable. ... View more