The cassandra-stress tool 

A Java-based stress testing utility for basic benchmarking and load testing a Cassandra cluster.

The cassandra-stress tool is a Java-based stress testing utility for basic benchmarking and load testing a Cassandra cluster.

Data modeling choices can greatly affect application performance. Significant load testing over several trials is the best method for discovering issues with a particular data model. The cassandra-stress tool is an effective tool for populating a cluster and stress testing CQL tables and queries. Use cassandra-stress to:
  • Quickly determine how a schema performs.
  • Understand how your database scales.
  • Optimize your data model and settings.
  • Determine production capacity.
The cassandra-stress tool also supports a YAML-based profile for defining specific schemas with various compaction strategies, cache settings, and types. Sample files are located in:
  • Cassandra package installations: /usr/share/docs/cassandra/examples
  • Cassandra tarball installations: install_location/tools/
The YAML file supports user-defined keyspace, tables, and schema. The YAML file can be used to design tests of reads, writes, and mixed workloads.

When started without a YAML file, cassandra-stress creates a keyspace, keyspace1, and tables, standard1 or counter1, depending on what type of table is being tested. These elements are automatically created the first time you run a stress test and reused on subsequent runs. You can drop keyspace1 using DROP KEYSPACE. You cannot change the default keyspace and tables names without using a YAML file.

Usage:
  • DataStax Enterprise 5.0 Installer Services and package installations:
    $ cassandra-stress command [options]
  • DataStax Enterprise 5.0 Installer No-Services and tarball installations:
    $ cd install_location/resources/cassandra/tools
    $ bin/cassandra-stress command [options]
  • Cassandra package installations:
    $ cassandra-stress command [options]
  • Cassandra tarball installations:
    $ cd install_location/tools
    $ bin/cassandra-stress command [options]

cassandra-stress options 

Command Description
counter_read Multiple concurrent reads of counters. The cluster must first be populated by a counter_write test.
counter_write Multiple concurrent updates of counters.
help Display help: cassandra-stress help

Display help for an option: cassandra-stress help [options] For example: cassandra-stress help -schema

legacy Legacy support mode.
mixed Interleave basic commands with configurable ratio and distribution. The cluster must first be populated by a write test.
print Inspect the output of a distribution definition.
read Multiple concurrent reads. The cluster must first be populated by a write test.
user Interleave user provided queries with configurable ratio and distribution.
write Multiple concurrent writes against the cluster.
Important: Additional sub-options are available for each option in the following table. To get more detailed information on any of these, enter:
$ cassandra-stress help option

When entering the help command, be sure to precede the option name with a hyphen, as shown.

Cassandra-stress sub-options 

Sub-option Description
-col Column details, such as size and count distribution, data generator, names, and comparator.
Usage:
-col names=? [slice] [super=?] [comparator=?] [timestamp=?] [size=DIST(?)]
 or 
-col [n=DIST(?)] [slice] [super=?] [comparator=?] [timestamp=?] [size=DIST(?)]
-errors How to handle errors when encountered during stress testing.
Usage:
-errors [retries=?] [ignore]
-graph Graph results of cassandra-stress tests. Multiple tests can be graphed together.
Usage:
-graph file=? [revision=?] [title=?] [op=?]
-insert Insert specific options relating to various methods for batching and splitting partition updates.
Usage:
-insert [revisit=DIST(?)] [visits=DIST(?)] partitions=DIST(?) [batchtype=?] select-ratio=DIST(?) row-population-ratio=DIST(?)
-log Where to log progress and the interval to use.
Usage:
-log [level=?] [no-summary] [file=?] [interval=?]
-mode Thrift or CQL with options.
Usage:
-mode thrift [smart] [user=?] [password=?]
  or 
-mode native [unprepared] cql3 [compression=?] [port=?] [user=?] [password=?] [auth-provider=?] [maxPending=?] [connectionsPerHost=?]
  or
-mode simplenative [prepared] cql3 [port=?]
-node Nodes to connect to.
Usage:
-node [whitelist] [file=?]
-pop Population distribution and intra-partition visit order.
Usage:
-pop seq=? [no-wrap] [read-lookback=DIST(?)] [contents=?]
  or
-pop [dist=DIST(?)] [contents=?]
-port Specify port for connecting Cassandra nodes. Port can be specified for Cassandra native protocol, Thrift protocol or a JMX port for retrieving statistics.
Usage:
-port [native=?] [thrift=?] [jmx=?]
-rate Set the rate using the following options:
-rate threads=N [throttle=N] [fixed=N]
where
  • threads=N number of clients to run concurrently.
  • throttle=N throttle operations per second across all clients to a maximum rate (or less) with no implied schedule. Default is 0.
  • fixed=N expect fixed rate of operations per second across all clients with implied schedule. Default is 0.
OR
-rate [threads>=N] [threads<=N] [auto]
Where
  • threads>=N run at least this many clients concurrently. Default is 4.
  • threads<=N run at most this many clients concurrently. Default is 1000.
  • auto stop increasing threads once throughput saturates.
-sample Specify the number of samples to collect for measuring latency.
Usage:
-sample [history=?] [live=?] [report=?]
-schema Replication settings, compression, compaction, and so on.
Usage:
-schema [replication(?)] [keyspace=?] [compaction(?)] [compression=?]
-sendto Specify a stress server to send this command to.
Usage:
-sendToDaemon <host>
-transport Custom transport factories.
Usage:
-transport [factory=?] [truststore=?] [truststore-password=?] [ssl-protocol=?] [ssl-alg=?] [store-type=?] [ssl-ciphers=?]

Additional command-line parameters can modify how cassandra-stress runs:

Additional cassandra-stress parameters 

Command Description
cl=? Set the consistency level to use during cassandra-stress. Options are ONE, QUORUM, LOCAL_QUORUM, EACH_QUORUM, ALL, and ANY. Default is LOCAL_ONE.
clustering=DIST(?) Distribution clustering runs of operations of the same kind.
duration=? Specify the time to run, in seconds, minutes or hours.
err<? Specify a standard error of the mean; when this value is reached, cassandra-stress will end. Default is 0.02.
n>? Specify a minimum number of iterations to run before accepting uncertainly convergence.
n<? Specify a maximum number of iterations to run before accepting uncertainly convergence.
n=? Specify the number of operations to run.
no-warmup Do not warmup the process, do a cold start.
ops(?) Specify what operations to run and the number of each. (only with the user option)
profile=? Designate the YAML file to use with cassandra-stress. (only with the user option)
truncate=? Truncate the table created during cassandra-stress. Options are never, once, or always. Default is never.

Simple read and write examples 

# Insert (write) one million rows
$ cassandra-stress write n=1000000 -rate threads=50

# Read two hundred thousand rows.
$ cassandra-stress read n=200000 -rate threads=50

# Read rows for a duration of 3 minutes.
$ cassandra-stress read duration=3m -rate threads=50

# Read 200,000 rows without a warmup of 50,000 rows first.
$ cassandra-stress read n=200000 no-warmup -rate threads=50

View schema help 

$ cassandra-stress help -schema
replication([strategy=?][factor=?][<option 1..N>=?]):                 Define the replication strategy and any parameters
    strategy=? (default=org.apache.cassandra.locator.SimpleStrategy)  The replication strategy to use
    factor=? (default=1)                                              The number of replicas
keyspace=? (default=keyspace1)                                        The keyspace name to use
compaction([strategy=?][<option 1..N>=?]):                            Define the compaction strategy and any parameters
    strategy=?                                                        The compaction strategy to use
compression=?                                                         Specify the compression to use for SSTable, default:no compression

Populate the database 

Generally it is easier to let cassandra-stress create the basic schema and then modify it in CQL:

#Load one row with default schema
$ cassandra-stress write n=1 cl=one -mode native cql3 -log file=create_schema.log
 
#Modify schema in CQL
$ cqlsh
 
#Run a real write workload
$ cassandra-stress write n=1000000 cl=one -mode native cql3 -schema keyspace="keyspace1" -log file=load_1M_rows.log

Change the replication strategy 

Changes the replication strategy to NetworkTopologyStrategy and targets one node named existing.

$ cassandra-stress write n=500000 no-warmup -node existing -schema "replication(strategy=NetworkTopologyStrategy, existing=2)"

Run a mixed workload 

When running a mixed workload, you must escape parentheses, greater-than and less-than signs, and other such things. This example invokes a workload that is one-quarter writes and three-quarters reads.

$ cassandra-stress mixed ratio\(write=1,read=3\) n=100000 cl=ONE -pop dist=UNIFORM\(1..1000000\) -schema keyspace="keyspace1" -mode native cql3 -rate threads\>=16 threads\<=256 -log file=~/mixed_autorate_50r50w_1M.log

Notice the following in this example:

  1. The ratio parameter requires backslash-escaped parenthesis.
  2. The value of n used in the read phase is different from the value used in write phase. During the write phase, n records are written. However in the read phase, if n is too large, it is inconvenient to read all the records for simple testing. Generally, n does not need be large when validating the persistent storage systems of a cluster.

    The -pop dist=UNIFORM\(1..1000000\) portion says that of the n=100,000 operations, select the keys uniformly distributed between 1 and 1,000,000. Use this when you want to specify more data per node than what fits in DRAM.

  3. In the rate section, the greater-than and less-than signs are escaped. If not escaped, the shell attempts to use them for IO redirection: the shell tries to read from a non-existent file called =256 and create a file called =16. The rate section tells cassandra-stress to automatically attempt different numbers of client threads and not test less that 16 or more than 256 client threads.

Standard mixed read/write workload keyspace for a single node 

CREATE KEYSPACE "keyspace1" WITH replication = {
  'class': 'SimpleStrategy',
  'replication_factor': '1'
};
USE "keyspace1";
CREATE TABLE "standard1" (
  key blob,
  "C0" blob,
  "C1" blob,
  "C2" blob,
  "C3" blob,
  "C4" blob,
  PRIMARY KEY (key)
) WITH
  bloom_filter_fp_chance=0.010000 AND
  caching='KEYS_ONLY' AND
  comment='' AND
  dclocal_read_repair_chance=0.000000 AND
  gc_grace_seconds=864000 AND
  index_interval=128 AND
  read_repair_chance=0.100000 AND
  replicate_on_write='true' AND
  default_time_to_live=0 AND
  speculative_retry='99.0PERCENTILE' AND
  memtable_flush_period_in_ms=0 AND
  compaction={'class': 'SizeTieredCompactionStrategy'} AND
  compression={'class': 'LZ4Compressor'};

Split up a load over multiple cassandra-stress instances on different nodes 

This example demonstrates loading into large clusters, where a single cassandra-stress load generator node cannot saturate the cluster. In this example, $NODES is a variable whose value is a comma delimited list of IP addresses such as 10.0.0.1, 10.0.0.2, and so on.

#On Node1
$ cassandra-stress write n=1000000 cl=one -mode native cql3 -schema keyspace="keyspace1" -pop seq=1..1000000 -log file=~/node1_load.log -node $NODES
 
#On Node2
$ cassandra-stress write n=1000000 cl=one -mode native cql3 -schema keyspace="keyspace1" -pop seq=1000001..2000000 -log file=~/node2_load.log -node $NODES 

Run cassandra-stress with authentication 

The following example shows using the -mode option to supply a username and password:

$ cassandra-stress -mode native cql3 user=cassandra password=cassandra no-warmup cl=QUORUM
Note: Check the documentation of the transport option for SSL authentication.

Run cassandra-stress with authentication and SSL encryption

The following example shows using the -mode option to supply a username and password, and the -transportation option for SSL parameters:

$ cassandra-stress write n=100k cl=ONE no-warmup -mode native cql3 user=cassandra password=cassandra 
-transport truststore=/usr/local/lib/dsc-cassandra/conf/server-truststore.jks truststore-password=truststorePass 
factory=org.apache.cassandra.thrift.SSLTransportFactory 
keystore=/usr/local/lib/dsc-cassandra/conf/server-keystore.jks keystore-password=myKeyPass
Note: Cassandra authentication and SSL encryption must already be configured before executing cassandra-stress with these options. The example shown above uses self-signed CA certificates.

Run cassandra-stress using the truncate option 

This option must be inserted before the mode option, otherwise the cassandra-stress tool won't apply truncation as specified.

The following example shows the truncate command:

$ cassandra-stress write n=100000000 cl=QUORUM truncate=always -schema keyspace=keyspace-rate threads=200 -log file=write_$NOW.log

Use a YAML file to run cassandra-stress 

This example uses a YAML file named cqlstress-example.yaml, which contains the keyspace and table definitions, and a query definition. The keyspace name and definition are the first entries in the YAML file:
keyspace: perftesting

keyspace_definition: 

  CREATE KEYSPACE perftesting WITH replication = { 'class': 'SimpleStrategy', 'replication_factor': 3};
The table name and definition are created in the next section using CQL:

table: users

table_definition:

  CREATE TABLE users (
    username text,
    first_name text,
    last_name text,
    password text,
    email text,
    last_access timeuuid,
    PRIMARY KEY(username)
  );
In the extra_definitions section you can add secondary indexes or materialized views to the table:


extra_definitions:
  - CREATE MATERIALIZED VIEW perftesting.users_by_first_name AS SELECT * FROM perftesting.users WHERE first_name IS NOT NULL and username IS NOT NULL PRIMARY KEY (first_name, username);
  - CREATE MATERIALIZED VIEW perftesting.users_by_first_name2 AS SELECT * FROM perftesting.users WHERE first_name IS NOT NULL and username IS NOT NULL PRIMARY KEY (first_name, username);
  - CREATE MATERIALIZED VIEW perftesting.users_by_first_name3 AS SELECT * FROM perftesting.users WHERE first_name IS NOT NULL and username IS NOT NULL PRIMARY KEY (first_name, username);
The population distribution can be defined for any column in the table. This section specifies a uniform distribution between 10 and 30 characters for username values in gnerated rows, that the values in the generated rows willcreates , a uniform distribution between 20 and 40 characters for generated startdate over the entire Cassandra cluster, and a Gaussian distribution between 100 and 500 characters for description values.
columnspec:
  - name: username
    size: uniform(10..30)
  - name: first_name
    size: fixed(16)
  - name: last_name
    size: uniform(1..32)
  - name: password
    size: fixed(80) # sha-512
  - name: email
    size: uniform(16..50)
  - name: startdate
    cluster: uniform(20...40)
  - name: description
    size: gaussian(100...500)
After the column specifications, you can add specifications for how each batch runs. In the following code, the partitions value directs the test to use the column definitions above to insert a fixed number of rows in the partition in each batch:
insert:
  partitions: fixed(10)
  batchtype: UNLOGGED
The last section contains a query, read1, that can be run against the defined table.
queries:
  read1:
    cql: select * from users where username = ? and startdate = ?
    fields: samerow     # samerow or multirow (select arguments from the same row, or randomly from all rows in the partition)
The following example shows using the user option and its parameters to run cassandra-stress tests from cqlstress-example.yaml:
$ cassandra-stress user profile=tools/cqlstress-example.yaml n=1000000 ops\(insert=3,read1=1\) no-warmup cl=QUORUM
Notice that:
  • The user option is required for the profile and opt parameters.
  • The value for the profile parameter is the path and filename of the .yaml file.
  • In this example, -n specifies the number of batches that run.
  • The values supplied for ops specifies which operations run and how many of each. These values direct the command to insert rows into the database and run the read1 query.

    How many times? Each insert or query counts as one batch, and the values in ops determine how many of each type are run. Since the total number of batches is 1,000,000, and ops says to run three inserts for each query, the result will be 750,000 inserts and 250,000 of the read1 query.

    Use escaping backslashes when specifying the ops value.

For more information, see Improved Cassandra 2.1 Stress Tool: Benchmark Any Schema – Part 1.

Use the -graph option 

In Cassandra 3.2 and later, the -graph option provides visual feedback for cassandra-stress tests. A file must be named to build the resulting HTML file. A title and revision are optional, but revision must be used if multiple stress tests are graphed on the same output.
$ cassandra-stress user profile=tools/cqlstress-example.yaml ops\(insert=1\) -graph file=test.html title=test revision=test1 

An interactive graph can be displayed with a web browser: