Incremental Updates in HIVE Using SQOOP PDF

PROOF OF CONCEPT FOR SQOOP INCREMENTAL
AND
STEPS TO BE TAKEN TO EXECUTE SQOOP JOBS ON OOZIE
- created by [email protected]
This document attempts to explain how sqoop can be used to provide upserts (update and inserts) in hive from
mysql database. Additionally, this document also explains how the entire process can be automated by writing
workflow in oozie.
Certain precautions are needed to be observed when executing sqoop jobs in oozie. These are noted down at
the end.
Dataset for this document has been borrowed from www.kaggle.com. Please refer database.csv in spacex-
missions.zip. First, we will load the data into mysql and perform some initial processing like adding timestamp
and primary key
1. Create database and table in mysql

CREATE DATABASE spacex; USE spacex;
CREATE TABLE spacex_missions (

Flight_Number VARCHAR(10),
Launch_Date VARCHAR(20),Launch_Time VARCHAR(10),Launch_Site VARCHAR(40),Vehicle_Type
VARCHAR(50),
Payload_Name VARCHAR(50),Payload_Type VARCHAR(50),Payload_Mass_kg
VARCHAR(15),Payload_Orbit VARCHAR(50),
Customer_Name VARCHAR(40),Customer_Type VARCHAR(40),Customer_Country VARCHAR(40),
Mission_Outcome VARCHAR(40),Failure_Reason VARCHAR(40),
Landing_Type VARCHAR(40),Landing_Outcome VARCHAR(40));
2. Load data into the table

LOAD DATA INFILE '/home/cloudera/Desktop/ajinkya/datasets/spacex-
missions/database.csv'
INTO TABLE spacex_missions
FIELDS TERMINATED BY ',' OPTIONALLY ENCLOSED BY '"'
LINES TERMINATED BY '\n'
IGNORE 1 LINES;
3. Create working table

CREATE TABLE spacex_missions_w AS
SELECT Flight_Number,
STR_TO_DATE(Concat(Launch_Date," ",Launch_Time),'%d %M %Y %H:%i') as
Launch_DateTime, Launch_Site, Vehicle_Type, Payload_Name, Payload_Type,
CAST(Payload_Mass_kg AS DECIMAL(10,3)) as Payload_Mass_kg,
Payload_Orbit, Customer_Name, Customer_Type, Customer_Country, Mission_Outcome,
Failure_Reason, Landing_Type, Landing_Outcome FROM spacex_missions
4. Add timestamp and primarykey to the table

ALTER TABLE spacex_missions_w
ADD COLUMN pid INT(5) UNSIGNED PRIMARY KEY AUTO_INCREMENT;
ALTER TABLE spacex_missions_w
add column Record_TS timestamp default current_timestamp on update
current_timestamp;
KNOWN BUG: If Record_TS is not populated with current_timestamp and instead is

populated with '0000', truncate spacex_missions_w and write
INSERT INTO spacex_missions_w

STR_TO_DATE(Concat(Launch_Date," ",Launch_Time),'%d %M %Y %H:%i') as
Launch_DateTime, Launch_Site, Vehicle_Type, Payload_Name, Payload_Type,
CAST(Payload_Mass_kg AS DECIMAL(10,3)) as Payload_Mass_kg,
Payload_Orbit, Customer_Name, Customer_Type, Customer_Country, Mission_Outcome,
Failure_Reason, Landing_Type, Landing_Outcome,
00, CURRENT_TIMESTAMP FROM spacex_missions
'spacex_missions_w' is our working table in mysql; we will create its replica in hive. We will be
constantly inserting and updating this table in mysql. After using four step incremental approach, these
changes will be reflected in Hive.
6. Replica of mysql table in hive
sqoop import \
--connect jdbc:mysql://10.170.245.155:3306/spacex \
--driver com.mysql.jdbc.Driver \
--username root \
--password cloudera \
--table spacex_missions_w \
--hive-import --create-hive-table \
--hive-table spacex.missions_base -m 1
In case we encounter '0000-00-00 00:00:00' values in timestamp column then we can

use jdbc:mysql://10.170.245.155:3306/spacex?zeroDateTimeBehavior=convertToNull to
convert into to null
7. Update mysql data
Now we make changes in mysql table worldbase and load only the updated data into hive using sqoop
incremental option. Data in mysql:
UPDATE spacex_missions_w SET customer_name='NA' WHERE pid=2023, pid=2025;
Updated mysql:
We can see record_ts has been updated by 20mins for updated columns. We will use that time in sqoop
incremental import.
8. Create incremental table
In order to support an on-going reconciliation between current records in HIVE and new change records,
two tables should be defined: missions_base and missions_inc
sqoop job \
--create spacex_import_job -- import \
--connect jdbc:mysql://quickstart.cloudera:3306/spacex \
--driver com.mysql.jdbc.Driver --username root \
--password cloudera --table spacex_missions_w \
--hive-import --create-hive-table --hive-table spacex.missions_inc -m 1 \
--check-column Record_TS --incremental lastmodified \
--last-value "2017-03-22 03:40:00"
sqoop job --exec spacex_import_job
We don't need to reset last-value while running the job next time as sqoop implicitly sets it.
9. Check data in hive

In hive we have two tables missions_base which contains original data and missions_inc which contains
updated data.
10. Reconcile View
This view combines record sets from both the Base (base_table) and Change (incremental_table) tables
and is reduced only to the most recent records for each unique “id”.
Create reconcile view of missions_base and missions_inc
CREATE VIEW missions_reconcile_view AS

SELECT t1.* FROM
(SELECT * FROM missions_base
UNION ALL
SELECT * from missions_inc) t1
JOIN
(SELECT pid, max(record_ts) max_ts FROM
(SELECT * FROM missions_base
UNION ALL
SELECT * from missions_inc) t_temp
GROUP BY pid) t2
ON t1.pid = t2.pid AND t1.record_ts = t2.max_ts;
11. Compaction of Data

The reconcile_view now contains the most up-to-date set of records and is now synchronized with
changes from the RDBMS source system. Before creating this table, any previous instances of the table
should be dropped as in the example below.
DROP TABLE IF EXISTS missions_reporting;

CREATE TABLE missions_reporting AS
SELECT * FROM missions_reconcile_view;
12. Purging incremental table data and reloading data into base table
DROP TABLE IF EXISTS missions_inc;
hdfs fs –rm –r /user/hive/warehouse/spacex.db/missions_inc
DROP TABLE missions_base;
CREATE TABLE missions_base LIKE missions_reporting
INSERT OVERWRITE TABLE missions_base SELECT * FROM missions_reporting;
This entire process can be automated by designing a oozie workflow. We will desgin oozie
workflow in Hue.
At the time of writing this document, there are three strategies known so as to automate incremental updates on
oozie.
1) Using SQOOP metastore

When SQOOPing updated data from saved job into HIVE, it was observed that SQOOP automatically
updated timestamp parameter for --last-value for next incremental import.
Now we may ask this question where does SQOOP store this timestamp?
The answer to this question is sqoop metastore. A particular SQOOP job is stored on a metastore of a particular
node. There is a difference between invoking SQOOP command via CLI on an edgenode and invoking SQOOP
via OOZIE. OOZIE may invoke SQOOP commands on any node and not necessarily on the node which has
SQOOP job and metastore saved. This problem can be tackled by using --meta-store option in SQOOP
command. This sets a single meta-store to be used for OOZIE. Additional research may be needed.
2) Calculating latest timestamp using Hive Query

In this strategy, before SQOOPing data, we calculate the latest timestamp of base_table in HIVE using
SELECT MAX() query. The output is echoed in shell. And by using capture-output facility of OOZIE we directly
pass this value into --last-value of SQOOP command. We will be using this strategy in this document.
3) Using additional column “process_date”

Review needed.
Designing the DAG on OOZIE:

Description of columns present in the scripts below-
record_ts : timestamp of the record
pid : primary key
1) Shell Action
Execute IMPALA query and echo the output
if [ "$(whoami)" = "yarn" ]; then

export USER=yarn
export PYTHON_EGG_CACHE=/tmp/impala-shell-python-egg-cache-${USER}
fi
Q0=$(impala-shell -i "localhost" -q "invalidate metadata;" )
echo "$Q0"
QUERY=$(impala-shell -i "localhost" -q "select max(record_ts) from

spacex.missions_base;")
Q1=`echo "$QUERY" | egrep -o "[0-9][0-9][0-9][0-9]-[0-2][0-9]-[0-3][0-9] [0-2][0-9]:
[0-6][0-9]:[0-9][0-9]"`
echo "MAXTS=$Q1"
Explanation of script:
We are setting environment variables to yarn because OOZIE executes its actions as YARN and not as
CLOUDERA. There may be a conflict between whoami and $USER.
[cloudera@quickstart ~]$ sudo -u yarn whoami

yarn
[cloudera@quickstart ~]$ whoami
cloudera
[cloudera@quickstart ~]$ sudo -u yarn echo $USER
cloudera
[cloudera@quickstart ~]$ echo $USER
cloudera
Secondly, we need to set up PYTHON_EGG_CACHE as Impala is written in Python and

might need a temporary cache directory.
Possible enhancement: invalidate metadata might be replaced by refresh as refresh is less

resource consuming and faster.
Query output:
+-----------------------+
| max(record_ts) |
+-----------------------+
| 2017-04-07 03:14:29.0 |
+-----------------------+
Therefore we use a regex expression to extract timestamp.
And echo the output as
MAXTS=2017-04-07 03:14:29.0
2) SQOOP Action
All the options will have to be entered as argument in Sqoop action
import
--connect
jdbc:mysql://quickstart.cloudera:3306/spacex
--username
root
--password
cloudera
--table
spacex_missions_w
-m
1
--check-column
Record_TS
--incremental
lastmodified
--last-value
${wf:actionData("shell-1860")["MAXTS"]}
--merge-key
pid
--split-by
pid
--target-dir
/user/cloudera/hive/external/missions_inc
Files selected under SQOOP action

/user/cloudera/jars/mysql-connector-java-5.1.34-bin.jar
/user/cloudera/hive/hive-site.xml
It must be noted that create-hive-table doesn't work with OOZIE. However, hive-import may
work and we can use this facility. However, here we have uploaded data into HDFS and we
will create an external table on top of it. Specifying, --merge-key primary_key is
required so as to execute incremental update the second time. (Meaning first incremental
update may work without specifying merge-key). Everything else is same as SQOOP command
mentioned in Hortonworks 4 Steps Incrmental Guide.
3) HIVE Action to create incremental table
create external table spacex.missions_inc (

flight_number string,
launch_datetime string,
launch_site string,
vehicle_type string,
payload_name string,
payload_type string,
payload_mass_kg double,
payload_orbit string,
customer_name string,
customer_type string,
customer_country string,
mission_outcome string,
failure_reason string,
landing_type string,
landing_outcome string,
pid bigint,
record_ts string)
ROW FORMAT delimited
FIELDS TERMINATED BY ','
STORED AS TEXTFILE
LOCATION '/user/cloudera/hive/external/missions_inc/';
4) HIVE Action to reconcile and purge
--RECONCILE AND PURGE HIVE TABLES

DROP TABLE IF EXISTS spacex.missions_reporting;
CREATE TABLE spacex.missions_reporting AS
SELECT * FROM spacex.missions_reconcile_view;
DROP TABLE IF EXISTS spacex.missions_inc;
--DROP TABLE IF EXISTS spacex.missions_base;
--CREATE TABLE spacex.missions_base LIKE spacex.missions_reporting;
INSERT OVERWRITE TABLE spacex.missions_base SELECT * FROM spacex.missions_reporting;
Instead of deleting Base table, it might be a good idea to overwrite it.
5) HDFS Action
Possible Enhancement: All the hive actions mentioned above can be executed in IMPALA and
performance can be improved especially those which involve JOINS

Incremental Updates in HIVE Using SQOOP PDF

Uploaded by

Copyright:

Available Formats

Incremental Updates in HIVE Using SQOOP PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Incremental Updates in HIVE Using SQOOP PDF

Uploaded by

Copyright:

Available Formats

PROOF OF CONCEPT FOR SQOOP INCREMENTAL

1. Create database and table in mysql

CREATE TABLE spacex_missions (

2. Load data into the table

3. Create working table

4. Add timestamp and primarykey to the table

KNOWN BUG: If Record_TS is not populated with current_timestamp and instead is

INSERT INTO spacex_missions_w

6. Replica of mysql table in hive

In case we encounter '0000-00-00 00:00:00' values in timestamp column then we can

7. Update mysql data

8. Create incremental table

sqoop job --exec spacex_import_job

9. Check data in hive

CREATE VIEW missions_reconcile_view AS

11. Compaction of Data

DROP TABLE IF EXISTS missions_reporting;

1) Using SQOOP metastore

2) Calculating latest timestamp using Hive Query

3) Using additional column “process_date”

Designing the DAG on OOZIE:

if [ "$(whoami)" = "yarn" ]; then

QUERY=$(impala-shell -i "localhost" -q "select max(record_ts) from

[cloudera@quickstart ~]$ sudo -u yarn whoami

Secondly, we need to set up PYTHON_EGG_CACHE as Impala is written in Python and

Possible enhancement: invalidate metadata might be replaced by refresh as refresh is less

All the options will have to be entered as argument in Sqoop action

Files selected under SQOOP action

3) HIVE Action to create incremental table

create external table spacex.missions_inc (

4) HIVE Action to reconcile and purge

--RECONCILE AND PURGE HIVE TABLES

Instead of deleting Base table, it might be a good idea to overwrite it.

You might also like