Skip to content

Latest commit

 

History

History
54 lines (44 loc) · 1.47 KB

common-issues-during-cluster-upgrade.md

File metadata and controls

54 lines (44 loc) · 1.47 KB
title description services documentationcenter author manager ms.service ms.custom ms.devlang ms.topic ms.tgt_pltfrm ms.workload ms.date ms.author
Common issues during cluster upgrade | Microsoft Docs
Diagnosing and fixing issue during cluster upgrade
hdinsight
duoxu
ashitg
hdinsight
hdinsightactive
na
article
na
big-data
11/10/2017
duoxu

HDP upgrade causes no data in Phoenix views

We have a customer upgrading their cluster from 3.4 to 3.5 (HDP 2.4 to HDP 2.5), there is one view contains no data after new cluster was created.

i see the view in sqlline... i see no data in the view in sqlline though

0: jdbc:phoenix:zk1-profil> select count(*) from table1_view;
+-----------+
| COUNT(1)  |
+-----------+
| 1         |
+-----------+

however in the table that the view is backed by...

0: jdbc:phoenix:zk1-profil> select count(*) from table1;
+-----------+
| COUNT(1)  |
+-----------+
| 1823616   |
+-----------+

HBCK returned fine, so there are no offline regions.

The root cause is actually that index table for views (all indexes for view are stored in a single physical HBase table) is truncated during upgrade and it’s expected that user will recreate them.

In phoenix-core/src/main/java/org/apache/phoenix/util/UpgradeUtil.java, there are comments that

 // Disable the view index and truncate the underlying hbase table. 
 // Users would need to rebuild the view indexes. 

Users should drop and recreate all indexes after upgrade.