Cassandra

Data Modelling and

Queries with CQL3
By Markus Klems
(2013)
Source: http://geek-and-poke.com
Data Model
● Keyspace (Database)
● Column Family (Table)
● Keys and Columns
A column

column_name

value

timestamp

the timestamp field is

used by Cassandra for conflict
resolution: “Last Write Wins”
 Column family (Table)
partition key columns ...

email name tel

101
[email protected] otto 12345

email name tel tel2

103
[email protected] karl 6789 12233

name
104
linda
Table with standard PRIMARY KEY
CREATE TABLE messages (
msg_id timeuuid PRIMARY KEY,
author text,
body text
);
Table: Tweets
PRIMARY KEY
= msg_id

author body
9990
otto Hello World!

author body
9991
linda Hi, Otto
Table with compound PRIMARY KEY
CREATE TABLE timeline (
user_id uuid,
msg_id timeuuid,
author text,
body text,
PRIMARY KEY (user_id, msg_id)
);
 “Wide-row” Table: Timeline
PRIMARY KEY = user_id + msg_id

partition key column

msg_id author body

103
9990 otto Hello World!

msg_id author body

103
9991 linda Hi @otto
Timeline Table is partitioned by user
and locally clustered by msg

103
msg_id ... 211
msg_id ...
9990 ... 8090 ...
103
msg_id ... 211
msg_id ...
9994 ... 8555 ...
104
msg_id ... 211
msg_id ...
9881 ... 9678 ...
104
msg_id ... 212
msg_id ...
9999 ... 9877 ...

Node A Node B
 Comparison: RDBMS vs. Cassandra

RDBMS Data Design Cassandra Data Design

Users Table Users Table
user_id name email user_id name email

101 otto [email protected] 101 otto [email protected]

Tweets Table Tweets Table

tweet_id author_id body tweet_id author_id name body
9990 101 Hello! 9990 101 otto Hello!

Followers Table Follows Table Followed Table

follows_ followed
id user_id follows_list id followed_list
id _id
4321 104 101 104 [101,117] 101 [104,109]
Exercise: Launch 1 Cassandra Node

Data Center DC1 /etc/conf/cassandra.yaml

(Germany)
A~$ start service cassandra
seeds: A
rpc_address: 0.0.0.0
A

launch server A
Exercise: Start CQL Shell

Data Center DC1

(Germany)
A~$ cqlsh

A
Intro: CLI, CQL2, CQL3
● CQL is “SQL for Cassandra”
● Cassandra CLI deprecated, CQL2 deprecated
● CQL3 is default since Cassandra 1.2
$ cqlsh
● Pipe scripts into cqlsh
$ cat cql_script | cqlsh
● Source files inside cqlsh
cqlsh> SOURCE '~/cassandra_training/cql3/
01_create_keyspaces';
CQL3
● Create a keyspace
● Create a column family
● Insert data
● Alter schema
● Update data
● Delete data
● Apply batch operation
● Read data
● Secondary Index
● Compound Primary Key
● Collections
● Consistency level
● Time-To-Live (TTL)
● Counter columns
● sstable2json utility tool
Create a SimpleStrategy keyspace
● Create a keyspace with SimpleStrategy and
"replication_factor" option with value "3" like
this:

cqlsh> CREATE KEYSPACE <ksname>

WITH REPLICATION =
{'class':'SimpleStrategy',
'replication_factor':3};
Exercise: Create a SimpleStrategy
keyspace
● Create a keyspace "simpledb" with SimpleStrategy and
replication factor 1.
Exercise: Create a SimpleStrategy
keyspace

cqlsh> CREATE KEYSPACE simpledb

WITH REPLICATION = {
'class' : 'SimpleStrategy',
'replication_factor' : 1 };
cqlsh> DESCRIBE KEYSPACE simpledb;
Create a NetworkTopologyStrategy
keyspace
Create a keyspace with NetworkTopologyStrategy and
strategy option "DC1" with a value of "1" and "DC2" with a
value of "2" like this:

cqlsh> CREATE KEYSPACE <ksname>

WITH REPLICATION = {
'class':'NetworkTopologyStrategy',
'DC1':1,
'DC2':2
};
Exercise Create Table “users”
● Connect to the "twotter" keyspace.
cqlsh> USE twotter;
● Create new column family (Table) named "users".
cqlsh:twotter> CREATE TABLE users (
id int PRIMARY KEY,
name text,
email text
);
cqlsh:twotter> DESCRIBE TABLES;
cqlsh:twotter> DESCRIBE TABLE users;

*we use int instead of uuid in the exercises for the sake of readability
Exercise: Create Table "messages"
● Create a new Table named "messages" with the
attributes "posted_on", "user_id", "user_name", "body",
and a primary key that consists of "user_id" and
"posted_on".
cqlsh:twotter> CREATE TABLE messages (
posted_on bigint,
user_id int,
user_name text,
body text,
PRIMARY KEY (user_id, posted_on)
);
*we use bigint instead of timeuuid in the exercises for the sake of readability
Exercise: Insert data into Table
"users" of keyspace "twotter"
cqlsh:twotter>
INSERT INTO users(id, name, email)
VALUES (101, 'otto', '[email protected]');

cqlsh:twotter> ... insert more records ...

cqlsh> SOURCE
'~/cassandra_training/cql3/03_insert';
Exercise: Insert message records

cqlsh:twotter>
INSERT INTO messages (user_id, posted_on,
user_name, body)
VALUES (101, 1384895178, 'otto', 'Hello!');

cqlsh:twotter> SELECT * FROM messages;

Read data
cqlsh:twotter> SELECT * FROM users;

id | email | name
-----+--------------+-------
105 | [email protected] | gerd
104 | [email protected] | linda
102 | null | jane
106 | [email protected] | heinz
101 | [email protected] | otto
103 | null | karl
Update data
cqlsh:twotter> UPDATE users
SET email = '[email protected]'
WHERE id = 102;

id | email | name
-----+----------------+-------
105 | [email protected] | gerd
104 | [email protected] | linda
102 | [email protected] | jane
106 | [email protected] | heinz
101 | [email protected] | otto
103 | null | karl
Delete data
● Delete columns
cqlsh:twotter> DELETE email
FROM users
WHERE id = 105;
● Delete an entire row

cqlsh:twotter> DELETE FROM users

WHERE id = 106;
Delete data

id | email | name
-----+----------------+-------
105 | null | gerd
104 | [email protected] | linda
102 | [email protected] | jane
.
101 | [email protected] | otto
103 | null | karl
Batch operation
● Execute multiple mutations with a single operation
cqlsh:twotter>
BEGIN BATCH
INSERT INTO users(id, name, email)
VALUES(107, 'john','[email protected]')
INSERT INTO users(id, name)
VALUES(108, 'michael')
UPDATE users
SET email = '[email protected]'
WHERE id = 108
DELETE FROM users WHERE id = 105
APPLY BATCH;
Batch operation

id | email | name
-----+----------------+---------
.
107 | [email protected] | john
108 | [email protected] | michael
104 | [email protected] | linda
102 | [email protected] | jane
101 | [email protected] | otto
103 | null | karl
Secondary Index
cqlsh:twotter>
CREATE INDEX name_index ON users(name);

cqlsh:twotter>
CREATE INDEX email_index ON users(email);

cqlsh:twotter> SELECT name, email FROM

users WHERE name = 'otto';

cqlsh:twotter> SELECT name, email FROM

users WHERE email = '[email protected]';
Alter Table Schema
cqlsh:twotter>
ALTER TABLE users ADD password text;

cqlsh:twotter>
ALTER TABLE users
ADD password_reset_token text;

* Given its flexible schema, Cassandra’s CQL ALTER finishes much

quicker than RDBMS SQL ALTER where all existing records need to be
updated.
Alter Table Schema
id | email | name | password | password_reset_token
-----+----------------+---------+----------+----------------------
107 | [email protected] | john | null | null
108 | [email protected] | michael | null | null
104 | [email protected] | linda | null | null
102 | [email protected] | jane | null | null
101 | [email protected] | otto | null | null
103 | null | karl | null | null
Collections - Set
CQL3 introduces collections for storing complex data
structures, namely the following: set, list, and map. This is
the CQL way of modelling many-to-one relationships.
1. Let us add a set of "hobbies" to the Table "users".
cqlsh:twotter> ALTER TABLE users ADD hobbies
set<text>;
cqlsh:twotter> UPDATE users SET hobbies =
hobbies +
{'badminton','jazz'} WHERE id = 101;
Collections - List
2. Now create a Table "followers" with a list of followers.
cqlsh:twotter> CREATE TABLE followers (
user_id int PRIMARY KEY,
followers list<text>);
cqlsh:twotter> INSERT INTO followers (
user_id, followers)
VALUES (101, ['willi','heinz']);

cqlsh:twotter> SELECT * FROM followers;

user_id | followers
---------+--------------------
101 | ['willi', 'heinz']
Collections - Map
3. Add a map to the Table "messages".
cqlsh:twotter>
ALTER TABLE messages
ADD comments map<text, text>;

cqlsh:twotter>
UPDATE messages
SET comments = comments + {'otto':'thx!'}
WHERE user_id = 103
AND posted_on = 1384895223;
Consistency Level
● Set the consistency level for all subsequent requests:
cqlsh:twotter> CONSISTENCY ONE;
cqlsh:twotter> CONSISTENCY QUORUM;
cqlsh:twotter> CONSISTENCY ALL;

● Show the current consistency level:

cqlsh:twotter> CONSISTENCY;
Exercise: Consistency Level
● Set the consistency level to ANY and execute a SELECT
statement.
Exercise: Consistency Level
● Set the consistency level to ANY and execute a SELECT
statement.

Bad Request: ANY ConsistencyLevel is only

supported for writes
Exercise: Time-To-Live (TTL)
● Insert a user record with a password reset
token with a 77 second TTL value.
cqlsh:twotter>
INSERT INTO users (id, name,
password_reset_token)
VALUES (109, 'timo', 'abc-xyz-123')
USING TTL 77;
Exercise: Time-To-Live (TTL)
● The INSERT statement before will delete the
entire user record after 77 seconds.
● This is what we actually wanted to do:
cqlsh:twotter> INSERT INTO users (id, name)
VALUES(110, 'anna');
cqlsh:twotter> UPDATE users USING TTL 77
SET password_reset_token =
'abc-xyz-123'
WHERE id = 110;
Time-To-Live (TTL)
● Check the TTL expiration time in seconds.

cqlsh:twotter> SELECT TTL

(password_reset_token)
FROM messages
WHERE user_id = 110;
Counter Columns
Create a Counter Column Table that counts
"upvote" and "downvote" events.

cqlsh:twotter> CREATE TABLE votes (

user_id int,
msg_created_on bigint,
upvote counter,
downvote counter,
PRIMARY KEY (user_id, msg_created_on)
);
Counter Columns
cqlsh:twotter>
UPDATE votes SET upvote = upvote + 1
WHERE user_id = 101
AND msg_created_on = 1234;

cqlsh:twotter>
UPDATE votes
SET downvote = downvote + 2
WHERE user_id = 101
AND msg_created_on = 1234;

cqlsh:twotter> SELECT * FROM votes;

sstable2json utility tool
$ sstable2json
var/lib/cassandra/data/twotter/users/*.db >
*.json

2X MB
.json file
X MB
.db file
Exercise: sstable2json
● Insert a few records
● Flush the users column family to disk and create a json
representation of a *.db file.
Solution: sstable2json
$ nodetool flush twotter users
$ sstable2json
/var/lib/cassandra/data/twotter/users/twotte
r-users-jb-1-Data.db > twotter-users.json
$ cat twotter-users.json
[{"key": "00000069","columns": [["","",
1384963716697000], ["email","[email protected]",
1384963716697000], ["name","gerd",
1384963716697000]]},
{"key": "00000068","columns": [["","",
1384963716685000], ...
CQL v3.1.0
(New in Cassandra 2.0)

● IF keyword
● Lightweight transactions (“Compare-And-Set”)
● Triggers (experimental!!)
● CQL paging support
● Drop column support
● SELECT column aliases
● Conditional DDL
● Index enhancements
● cqlsh COPY
IF Keyword

cqlsh> DROP KEYSPACE twotter;

cqlsh> DROP KEYSPACE twotter;
Bad Request: Cannot drop non existing
keyspace 'twotter'.
cqlsh> DROP KEYSPACE IF EXISTS twotter;
Lightweight Transactions
● Compare And Set (CAS)
● Example: without CAS, two users attempting
to create a unique user account in the same
cluster could overwrite each other’s work
with neither user knowing about it.

Source: http://www.datastax.com/documentation/cassandra/2.0/webhelp/cassandra/dml/dml_about_transactions_c.html
Lightweight Transactions
1. Register a new user
cqlsh:twotter>
INSERT INTO users (id, name, email)
VALUES (110, 'franz', '[email protected]')
IF NOT EXISTS;
2. Perform a CAS reset of Karl’s email.
cqlsh:twotter>
UPDATE users
SET email = '[email protected]'
WHERE id = 110
IF email = '[email protected]';
Exercise: Lightweight Transactions

● Perform a failing CAS e-mail reset:

cqlsh:twotter>
UPDATE users
SET email = '[email protected]'
...
Exercise: Lightweight Transactions

● Perform a failing CAS e-mail reset:

cqlsh:twotter>
UPDATE users
SET email = '[email protected]'
WHERE id = 110
IF email = '[email protected]';

[applied] | email
-----------+-------------
False | [email protected]
Exercise: Lightweight Transactions
● Write a password reset method by using an
expiring password_reset_token column and
a CAS password update query.
Exercise: Lightweight Transactions
cqlsh:twotter>
UPDATE users USING TTL 77
SET password_reset_token = 'abc-xyz-123'
WHERE id = 110;

cqlsh:twotter>
UPDATE users
SET password = 'geheim!'
WHERE id = 110
IF password_reset_token = 'abc-xyz-123';
Create a Trigger
(experimental feature)
● Triggers are written in Java.
● Triggers are currently an experimental feature
in Cassandra 2.0. Use with caution!

cqlsh:twotter>
CREATE TRIGGER myTrigger ON users USING 'org.
apache.cassandra.triggers.InvertedIndex'