Hi,

I recently upgraded Phoenix 4.2.2. to 4.3.1 on 9 node HDP 2.2 cluster.
HBase is configured with 6 region servers. Phoenix 4.2.2 was part of
Hortonworks distribution 2.2. I performed following steps to upgrade to
4.3.1 on all HBase nodes. After the upgrade, sqlline.py was successfully
connected to HBase and queries worked fine. However, bulk load utility
psql.py stopped working. I tried to load same CSV files into same HBase
table which were successfully uploaded in 4.2.2, but psql.py did not work
in 4.3.1. It simply displays the message "CSV Upsert complete. 0 rows
upserted" and does not return any error. I am able to successfully execute
.sql file using psql.py to create tables, but CSV file is not getting
loaded.


Steps I performed to upgrade 4.2.2 to 4.3.1 on all HBase nodes:


   - Phoenix was installed on HDP 2.2 as "yum install phoenix" on all 6
   region servers and HBase master. It created
   "/usr/hdp/2.2.0.0-2041/phoenix", "/usr/hdp/2.2.0.0-2041/phoenix/lib" and
   "/usr/hdp/2.2.0.0-2041/phoenix/bin" directories on all nodes.
   - In "/usr/hdp/2.2.0.0-2041/phoenix", there were two soft links created,
   one was phoenix-client.jar which was pointing
   to phoenix-4.2.0.2.2.0.0-2041-client.jar and another was phoenix-server.jar
   which was pointing to phoenix-4.2.0.2.2.0.0-2041-server.jar.
   - Copied phoenix-4.3.1-client.jar and phoenix-4.3.1-server.jar
   to "/usr/hdp/2.2.0.0-2041/phoenix", recreated two soft links to point to
   new 4.3.1 jars and renamed 4.2.2 jars to .bak.
   - Copied phoenix-core-4.3.1*.jar, phoenix-flume-4.3.1*.jar
   and phoenix-pig-4.3.1*.jar to "/usr/hdp/2.2.0.0-2041/phoenix/lib" and
   renamed corresponding 4.2.2 jars to .bak.
   - Copied 4.3.1 version of performance.py, phoenix_utils.py, psql.py
   and sqlline.py to "/usr/hdp/2.2.0.0-2041/phoenix/bin" and renamed 4.2.2
   version .py files to .bak.
   - Restarted HBase master and all 6 region servers.


I also upgraded 4.3.1 to 4.4.0 using similar steps but psql.py in 4.4.0
could not upload CSV file.

Any clue what might have gone wrong in the upgrade? Am I missing any jar
needed to run psql.py?


Thanks,

Hemal Parekh
Senior Data Warehouse Architect
 m. 240.449.4396
[image: Bitscopic Inc] <http://bitscopic.com>

Reply via email to