Skip to main content

Querying HBase Data with Impala via Hive’s HBaseStorageHandler

By Anonymous
This is a legacy but still useful walkthrough that shows how to expose HBase-resident data to Impala by going through Hive’s Metastore and the HBaseStorageHandler. Using US census ZIP code income data, we create an HBase table, map it with an external Hive table, bulk load the CSV data with Pig and finally query it from Impala. The pattern is mainly relevant today if you are keeping old CDH clusters alive or planning a migration away from Impala-on-HBase towards Parquet or Iceberg tables.

Note (2025): This article describes an older CDH/Impala/HBase pattern based on Hive’s HBaseStorageHandler. It is useful if you still maintain legacy Impala-on-HBase workloads or need to understand how such systems were wired. For new designs you will usually land data in Parquet or Iceberg tables and query them with Impala, Trino or Spark instead of reading directly from HBase.

Context: Impala, Hive Metastore and HBase

Impala uses the Hive Metastore Service to discover tables and their underlying storage. That means you can expose data sitting in HDFS, object storage or even HBase as regular tables and query them from Impala.

In this article we use Hive’s HBase handler (HBaseStorageHandler) to connect Hive and HBase, and then query the same data from Impala. If you have not seen this pattern before, think of it as:

  • HBase table → mapped as a Hive external table via HBaseStorageHandler
  • Hive external table → visible to Impala via the shared Hive Metastore
  • Impala → can run SQL against the HBase data with low latency, as long as the mapping is correct

In an older post I described how to connect Hive and HBase in more detail. That tutorial still applies to the basics of the HBase handler, so if you have never configured it before, read that first and then come back here.

Demo scenario: US census ZIP code incomes

For the demo we use public US census data (circa 2000) that contains ZIP codes and income values. The goal:

  1. Load the raw CSV into HBase.
  2. Expose the HBase table as a Hive external table via HBaseStorageHandler.
  3. Run queries against the data from Impala.

The scripts for this demo are available in my GitHub repository.

Sample CSV data

The raw CSV looks roughly like this:

cat DEC_00_SF3_P077_with_ann_noheader.csv

8600000US00601,00601,006015-DigitZCTA,0063-DigitZCTA,11102
8600000US00602,00602,006025-DigitZCTA,0063-DigitZCTA,12869
8600000US00603,00603,006035-DigitZCTA,0063-DigitZCTA,12423
8600000US00604,00604,006045-DigitZCTA,0063-DigitZCTA,33548
8600000US00606,00606,006065-DigitZCTA,0063-DigitZCTA,10603

Each row contains:

  • id: a unique key (we will use this as the HBase row key)
  • zip: ZIP code
  • desc1, desc2: descriptive fields
  • income: income value from the census dataset

Step 1 – Create the HBase table

Create a table in HBase with column families for each group of attributes:

create 'zipcode_hive', 'id', 'zip', 'desc', 'income'

In this example the row key is the census identifier (e.g. 8600000US00601), and we keep ZIP, descriptions and income in separate column families.

Step 2 – Expose HBase via a Hive external table

Next we create an external table in Hive that maps to the HBase table using HBaseStorageHandler and the hbase.columns.mapping property: 

CREATE EXTERNAL TABLE ZIPCODE_HBASE (
  key   STRING,
  zip   STRING,
  desc1 STRING,
  desc2 STRING,
  income STRING
)
ROW FORMAT DELIMITED
  FIELDS TERMINATED BY ','
STORED BY 'org.apache.hadoop.hive.hbase.HBaseStorageHandler'
WITH SERDEPROPERTIES (
  "hbase.columns.mapping" = ":key,zip:zip,desc:desc1,desc:desc2,income:income"
)
TBLPROPERTIES (
  "hbase.table.name" = "zipcode_hive"
);

Here we bind each Hive column to a corresponding HBase column:

  • :key → HBase row key
  • zip:zip → column zip in family zip
  • desc:desc1, desc:desc2 → two columns in the desc family
  • income:income → column income in family income

Step 3 – Load CSV data into HBase with Pig

We now copy the CSV file into HDFS and use Pig to populate the HBase table. You could also use a staging table in Hive, but Pig keeps the example compact. 

cat PopulateData.pig

copyFromLocal DEC_00_SF3_P077_with_ann_noheader.csv ziptest.csv

A = LOAD 'ziptest.csv'
    USING PigStorage(',')
    AS (id:chararray,
        zip:chararray,
        desc1:chararray,
        desc2:chararray,
        income:chararray);

STORE A INTO 'hbase://zipcode_hive'
  USING org.apache.pig.backend.hadoop.hbase.HBaseStorage(
    'zip:zip,desc:desc1,desc:desc2,income:income'
  );

The job takes a short time to run and then the data is available directly in HBase. A quick sanity check: 

scan 'zipcode_hive', LIMIT => 2

ROW                         COLUMN+CELL
8600000US00601             column=desc:desc1,   timestamp=..., value=006015-DigitZCTA
8600000US00601             column=desc:desc2,   timestamp=..., value=0063-DigitZCTA
8600000US00601             column=income:income,timestamp=..., value=11102
8600000US00601             column=zip:zip,      timestamp=..., value=00601

8600000US00602             column=desc:desc1,   timestamp=..., value=006025-DigitZCTA
8600000US00602             column=desc:desc2,   timestamp=..., value=0063-DigitZCTA
8600000US00602             column=income:income,timestamp=..., value=12869
8600000US00602             column=zip:zip,      timestamp=..., value=00602

Step 4 – Query the same data from Impala

Because Impala shares the Hive Metastore, the ZIPCODE_HBASE table we just created is immediately visible to Impala after you run INVALIDATE METADATA or an equivalent refresh.

A simple Impala session might look like this: 

-- connect to your Impala daemon
Using service name 'impala' for kerberos
Connected to hadoop1:21000
Server version: impalad version 1.0 RELEASE (build d1bf0d1dac339af3692ffa17a5e3fdae0aed751f)

INVALIDATE METADATA ZIPCODE_HBASE;

-- basic sanity query
SELECT * FROM ZIPCODE_HBASE LIMIT 4;

+----------------+------------------+----------------+--------+-------+
| key            | desc1            | desc2          | income | zip   |
+----------------+------------------+----------------+--------+-------+
| 8600000US00601 | 006015-DigitZCTA | 0063-DigitZCTA | 11102  | 00601 |
| 8600000US00602 | 006025-DigitZCTA | 0063-DigitZCTA | 12869  | 00602 |
| 8600000US00603 | 006035-DigitZCTA | 0063-DigitZCTA | 12423  | 00603 |
| 8600000US00604 | 006045-DigitZCTA | 0063-DigitZCTA | 33548  | 00604 |
+----------------+------------------+----------------+--------+-------+
Returned 4 row(s) in 0.42s

Example: simple income filter

Now we can run “normal” SQL against the census income data even though it lives in HBase: 

SELECT *
FROM ZIPCODE_HBASE
WHERE income BETWEEN '1000' AND '5000'
ORDER BY income DESC
LIMIT 20;

+----------------+------------------+----------------+--------+-------+
| key            | desc1            | desc2          | income | zip   |
+----------------+------------------+----------------+--------+-------+
| 8600000US64138 | 641385-DigitZCTA | 6413-DigitZCTA | 49995  | 64138 |
| 8600000US12477 | 124775-DigitZCTA | 1243-DigitZCTA | 49993  | 12477 |
| 8600000US33025 | 330255-DigitZCTA | 3303-DigitZCTA | 49991  | 33025 |
| 8600000US44119 | 441195-DigitZCTA | 4413-DigitZCTA | 49988  | 44119 |
| 8600000US34997 | 349975-DigitZCTA | 3493-DigitZCTA | 49982  | 34997 |
| 8600000US70665 | 706655-DigitZCTA | 7063-DigitZCTA | 49981  | 70665 |
| 8600000US28625 | 286255-DigitZCTA | 2863-DigitZCTA | 49981  | 28625 |
| 8600000US76134 | 761345-DigitZCTA | 7613-DigitZCTA | 49979  | 76134 |
| 8600000US44618 | 446185-DigitZCTA | 4463-DigitZCTA | 49978  | 44618 |
| 8600000US65714 | 657145-DigitZCTA | 6573-DigitZCTA | 49978  | 65714 |
| 8600000US77338 | 773385-DigitZCTA | 7733-DigitZCTA | 49976  | 77338 |
| 8600000US14622 | 146225-DigitZCTA | 1463-DigitZCTA | 49972  | 14622 |
| 8600000US84339 | 843395-DigitZCTA | 8433-DigitZCTA | 49972  | 84339 |
| 8600000US85020 | 850205-DigitZCTA | 8503-DigitZCTA | 49967  | 85020 |
| 8600000US64061 | 640615-DigitZCTA | 6403-DigitZCTA | 49964  | 64061 |
| 8600000US97361 | 973615-DigitZCTA | 9733-DigitZCTA | 49961  | 97361 |
| 8600000US30008 | 300085-DigitZCTA | 3003-DigitZCTA | 49960  | 30008 |
| 8600000US48634 | 486345-DigitZCTA | 4863-DigitZCTA | 49958  | 48634 |
| 8600000US47923 | 479235-DigitZCTA | 4793-DigitZCTA | 49946  | 47923 |
| 8600000US46958 | 469585-DigitZCTA | 4693-DigitZCTA | 49946  | 46958 |
+----------------+------------------+----------------+--------+-------+
Returned 20 row(s) in 1.08s

The important point is not the exact numbers but the pattern: once you have the Hive <-> HBase mapping in place, Impala can run SQL over your HBase data as if it were a regular table, while still reading from HBase at query time.

Where this pattern still makes sense (and where it doesn’t)

This Impala–Hive–HBase setup can still be useful when:

  • You maintain a legacy CDH cluster with HBase-backed tables that must stay online.
  • You need to troubleshoot or gradually migrate old Impala-on-HBase workloads.
  • You want a reference example for how HBaseStorageHandler was wired.

For new projects, you will usually:

  • Land raw data in HDFS or object storage,
  • Store it as Parquet or in an open table format like Iceberg,
  • Query it from engines such as Impala, Trino, Spark or Flink SQL,
  • Keep HBase (or other key-value stores) only for operational access paths, not as the main analytics store.

If you are in the process of modernizing such a stack, treat this article as a historical snapshot and migration reference, not as a recommended target architecture.

Related guides:

If you need help with distributed systems, backend engineering, or data platforms, check my Services.

Labels:

Most read articles

Why Is Customer Obsession Disappearing?

By
Many companies trade real customer-obsession for automated, low-empathy support. Through examples from Coinbase, PayPal, GO Telecommunications and AT&T, this article shows how reliance on AI chatbots, outsourced call centers, and KPI-driven workflows erodes trust, NPS and customer retention. It argues that human-centric support—treating support as strategic investment instead of cost—is still a core growth engine in competitive markets. It's wild that even with all the cool tech we've got these days, like AI solving complex equations and doing business across time zones in a flash, so many companies are still struggling with the basics: taking care of their customers. The drama around Coinbase's customer support is a prime example of even tech giants messing up. And it's not just Coinbase — it's a big-picture issue for the whole industry. At some point, the idea of "customer obsession" got replaced with "customer automation," and no...
Read more

What the Heck is Superposition and Entanglement?

By
This post is about superposition and interference in simple, intuitive terms. It describes how quantum states combine, how probability amplitudes add, and why interference patterns appear in systems such as electrons, photons and waves. The goal is to give a clear, non mathematical understanding of how quantum behavior emerges from the rules of wave functions and measurement. If you’ve ever heard the words superposition or entanglement thrown around in conversations about quantum physics, you may have nodded politely while your brain quietly filed them away in the "too confusing to deal with" folder.  These aren't just theoretical quirks; they're the foundation of mind-bending tech like Google's latest quantum chip, the Willow with its 105 qubits. Superposition challenges our understanding of reality, suggesting that particles don't have definite states until observed. This principle is crucial in quantum technologies, enabling phenomena like quantum comp...
Read more

SynthLink Compared to Google’s Natural Questions: A Practical Evaluation

By
SynthLink evaluates reasoning, synthesis and internal consistency across diverse question types. Google’s Natural Questions evaluates extractive QA: finding short text spans inside structured documents. Because real workloads require interpretation, abstraction and multi-step logic, SynthLink exposes capabilities and failure modes that NQ cannot measure. The two benchmarks are complementary, but SynthLink is more aligned with production tasks. Benchmarks such as Google’s Natural Questions (NQ) dominate model evaluation. They provide a reliable, academically stable test for extractive question answering: short queries, grounded answers, and constrained context ranges. But real workloads rarely look like NQ. Production systems must handle ambiguous inputs, multi-step reasoning, poorly structured prompts, and cases where no canonical answer exists. SynthLink was designed for this broader landscape. It focuses on evaluating reasoning, synthesis and internal consistency rather than snippe...
Read more