An easy management of CQL stress tests with the official 'cassandra-stress' tool. You can use the solution for tests not only cassandra, but also other solutions with support CQL such as Scylla, AstraDB, etc.
Key benefits:
- generate scripts for performance tests (with 'cassandra-stress') based on templates
- extract test summaries (from 'cassandra-stress' results)
- compare test results (as text tables and/or graphs)
- test cleaning/maintenance (ability to remove keyspace/table via CQL)
Pre-requisites:
- installed Python >= 3.11
- installed Java (java version based on cassandra-stress specification)
- installed 'cassandra-stress' tool (part of Apache Cassandra distribution)
- access to CQL solution (e.g. Cassandra, Scylla, Astra, etc.)
- open port 9042
- IP addresses for connection
- login information e.g. username/password
Motivation for this tool/repo:
- use the standard and verified official tool for testing CQL solutions
- focus on your own tests (not on creating a tool for performance tests)
- don't reinvent the wheel (unless necessary, the time is money)
You can see standard description:
python3.11 stress/stress_mng.py --helpUsage: stress_mng.py [OPTIONS] COMMAND [ARGS]...
Options:
--help Show this message and exit.
Commands:
compare Compare data from TXT(JSON) to the sub-dir 'graph'
extract Extract data from 'cassandra-stress' output to the sub-dir...
generate Generate performance tests as *.sh for 'cassandra-stress'
graph Create graphs from TXT(JSON) to the sub-dir 'graph'
remove Remove keyspace or table from CQL solution
version Print current version of key componentsYou can generate shell scripts for 'cassandra-stress', see:
python3.11 stress/stress_mng.py generate -e compareV4V5_separateTHR\_cass_*.env -l stress-cmd/_cass_all.shThe output ~140 calls with READ/WRITE tests and exact logic for thread growing (4, 8, 6, 32, ...)
#!/bin/sh
# GENERATED: 2024-12-05 16:23:20
curr_date=$(date +%Y-%m-%d_%H-%M-%S)
mkdir -p ./stress-output/$curr_date/
echo 'Based on: .\config\compareV4V5_separateTHR\_cass_v4_1_local_one.env'
echo 'START remove: ...'
python3.11 stress/stress_mng.py remove -e compareV4V5_separateTHR/_cass_v4_1_local_one.env -k keyspace1 -d stress -s 8
echo 'START write, 4x thread: 1/1...'
./apache-cassandra-5.0.2/tools/bin/cassandra-stress write duration=1m cl=LOCAL_ONE no-warmup -node 10.129.52.58,10.129.53.21,10.129.52.57 -mode user=perf password=perf prepared protocolVersion=4 connectionsPerHost=24 maxPending=384 -schema "replication(strategy=NetworkTopologyStrategy,factor=3)" "compaction(strategy=LeveledCompactionStrategy,sstable_size_in_mb=160,fanout_size=10)" -rate "threads=4" -reporting output-frequency=5s > "./stress-output/$curr_date/$curr_date v4 write_LOCAL_ONE_LCS_4xTHR.txt"
echo 'START write, 8x thread: 2/2...'
./apache-cassandra-5.0.2/tools/bin/cassandra-stress write duration=1m cl=LOCAL_ONE no-warmup -node 10.129.52.58,10.129.53.21,10.129.52.57 -mode user=perf password=perf prepared protocolVersion=4 connectionsPerHost=24 maxPending=384 -schema "replication(strategy=NetworkTopologyStrategy,factor=3)" "compaction(strategy=LeveledCompactionStrategy,sstable_size_in_mb=160,fanout_size=10)" -rate "threads=8" -reporting output-frequency=5s > "./stress-output/$curr_date/$curr_date v4 write_LOCAL_ONE_LCS_8xTHR.txt"
echo 'START write, 16x thread: 3/3...'
# ...python3.11 stress/stress_mng.py generate -e compareV4V5_sequenceTHR\_cass_*.env -l stress-cmd/_cass_seq_all.shThe output ~28 calls with READ/WRITE tests with gradual thread growing (till <=100 thread) based on default 'cassandra-stress' logic
#!/bin/sh
# GENERATED: 2024-12-05 16:24:10
curr_date=$(date +%Y-%m-%d_%H-%M-%S)
mkdir -p ./stress-output/$curr_date/
echo 'Based on: .\config\compareV4V5_sequenceTHR\_cass_v4_1_local_one.env'
echo 'START remove: ...'
python3.11 stress/stress_mng.py remove -e compareV4V5_sequenceTHR/_cass_v4_1_local_one.env -k keyspace1 -d stress -s 8
echo 'START write, 100x thread: 1/1...'
./apache-cassandra-5.0.2/tools/bin/cassandra-stress write duration=1m cl=LOCAL_ONE no-warmup -node 10.129.52.58,10.129.53.21,10.129.52.57 -mode user=perf password=perf prepared protocolVersion=4 connectionsPerHost=24 maxPending=384 -schema "replication(strategy=NetworkTopologyStrategy,factor=3)" "compaction(strategy=LeveledCompactionStrategy,sstable_size_in_mb=160,fanout_size=10)" -rate "threads<=100" -reporting output-frequency=5s > "./stress-output/$curr_date/$curr_date v4 write_LOCAL_ONE_LCS_100xTHR.txt"
echo 'START read, 100x thread: 2/1...'
./apache-cassandra-5.0.2/tools/bin/cassandra-stress read duration=1m cl=LOCAL_ONE no-warmup -node 10.129.52.58,10.129.53.21,10.129.52.57 -mode user=perf password=perf prepared protocolVersion=4 connectionsPerHost=24 maxPending=384 -rate "threads<=100" -reporting output-frequency=5s > "./stress-output/$curr_date/$curr_date v4 read_LOCAL_ONE_LCS_100xTHR.txt"
# ...The extract data from 'cassandra-stress' output in format CSV and TXT(JSON).
python3.11 stress/stress_mng.py extract -d "C:/Python/.NEW Compare V4 vs V5/FULLFinal/"The Compare data based on TXT(JSON) files to text (in console) or graphs (as PNG files)
python3.11 stress/stress_mng.py compare -d "C:/Python/.NEW Compare V4 vs V5/FULLFinal/"The Create graphs from TXT(JSON) to the sub-dir 'graph'.
python3.11 stress/stress_mng.py graph -d "C:/Python/.4NEW Compare V5 NONPROD/"It is useful for visual check, the inputs are TXT(JSON) files from extract command.
It is useful for table/excel compare (TAB as separator), the inputs are CSV files from extract command.
==== LOCAL_ONE===
Test case 4 8 16 24 36 54 81 4 8 16 24 36 54 81
v5 write_LOCAL_ONE_STCS 5938 11451 21774 29310 35638 39116 42557 0,7 0,7 0,7 0,8 1,0 1,4 1,9
v4 write_LOCAL_ONE_STCS 5874 11053 19690 26178 30213 28000 23091 0,7 0,7 0,8 0,9 1,2 1,9 3,4
...
==== LOCAL_QUORUM===
Test case 4 8 16 24 36 54 81 4 8 16 24 36 54 81
v5 write_LOCAL_QUORUM_STCS 3907 7058 12638 18065 23462 29958 32159 1,0 1,1 1,2 1,3 1,5 1,8 2,5
v4 write_LOCAL_QUORUM_STCS 3525 6394 10804 14313 18465 22709 25715 1,1 1,2 1,5 1,7 1,9 2,4 3,1
...The usage in the excel see a few final outputs:
