How to import compressed AVRO files to Impala table？

[#1]

In my work, I import AVRO files into impala tables by copy the files into HDFS then execute "refresh <table>" in impala.

But when I want to do it with compressed files, it didn't work. Here is what I do:

1. Enable Hive compress in hive shell:

hive> set hive.exec.compress.output=true;
hive> set avro.output.codec=bzip2;

2. Create a table:

CREATE TABLE `log_bzip2`(
  `timestamp` bigint COMMENT 'from deserializer',
  `appid` string COMMENT 'from deserializer',
  `clientid` string COMMENT 'from deserializer',
  `statkey` string COMMENT 'from deserializer',
  `expid` string COMMENT 'from deserializer',
  `modid` string COMMENT 'from deserializer',
  `value` double COMMENT 'from deserializer',
  `summary` string COMMENT 'from deserializer',
  `custom` string COMMENT 'from deserializer')
PARTITIONED BY (
  `day` string)
ROW FORMAT SERDE
  'org.apache.hadoop.hive.serde2.avro.AvroSerDe'
STORED AS INPUTFORMAT
  'org.apache.hadoop.hive.ql.io.avro.AvroContainerInputFormat'
OUTPUTFORMAT
  'org.apache.hadoop.hive.ql.io.avro.AvroContainerOutputFormat'
TBLPROPERTIES (
  'avro.schema.url'='hdfs://szq2.appadhoc.com:8020/user/hive/log.avsc');

3. Load the compressed AVRO file into HDFS:

hdfs dfs -put log.2016-03-07.1457184357726.avro.bz2 /user/hive/warehouse/adhoc_data_fast.db/log_bzip2/2016-03-07

4. Add partition and refresh in Impala shell:

alter table log_bzip2 add partition (day="2016-03-07") location '/user/hive/warehouse/adhoc_data_fast.db/log_bzip2/2016-03-07/';

refresh log_bzip2;

5. Query it but not work:

select * from log_bzip2 limit 10;
Query: select * from log_bzip2 limit 10

WARNINGS: Invalid AVRO_VERSION_HEADER: '42 5a 68 39 '

The only document I've found about enable compress with avro tables is this link: http://www.cloudera.com/documentation/archive/impala/2-x/2-1-x/topics/impala_avro.html#avro_compress... . 

How can I do it right? Thanks!

[#2]

Hi, couple of suggestions on attacking the problem by simplifying it somewhat:

1. Can you try with a table and data which is not partitioned?

2. Can you try with external table instead of managed, and use a path different to the hive warehouse?

Martin