README

Welcome to MariaDB QA's (mariadb-qa) implementation of the Squirrel fuzzer

The MariaDB QA team has derived signifcant value from the Squirrel fuzzer, and we would like to thank:
* The whole Squirrel team at https://github.com/s3team/Squirrel (also see whitepaper below)
  With special thanks to:
    * Changochen @ https://github.com/Changochen
    * Rui Zhong @ https://github.com/zr950624
* Shihao Wen

White paper: https://arxiv.org/pdf/2006.02398.pdf

The MariaDB QA team has added:
* Further automation like multi-fuzzer-nodes and multi-mariadb-instance management
* Removed the need to use docker instances
* Non-shared parallelization up to 10 nodes on one physical instance
* 20+ nodes possible with minor code/binary modifications
* Automatic UniqueID creation for all Crashing, ASAN, UBSAN, and TSAN bugs generated by the fuzzer
* Automatic deletion and handling of all already known/logged bugs
* Fuzzing server resource management (ref ~/ds and ~/memory as generated by mariadb-qa/linkit)
* Database server building and state management
* Multi-version, multi-vendor, multi build-type (debug/optimized) database server testing for testcases
* Automatic testcase reduction (and sampling as described in the last point), including sporadic issues
* Detailed automatic MariaDB JIRA bug report generation
* And more...

A number of these features were already present in the mariadb-qa framework, and these were modified to work with Squirrel

We also plan to release additional grammars in the future!

Continue your journey:

./SETUP: How to setup a server for mariadb-qa (pquery framework + fuzzer framework + more) based testing
         As this is quite change-intensive (changes many server settings) a dedicated instance is recommended
         A physical instance is recommended, or a high-end cloud instance like a dedidcated 30+ vCPUs, 120 GB Memory, SSD
         Minimum memory is 64GB, but to run with a reasonable level of reliability and speed, think 96GB+, SSD, 20 threads+
         OS: Ubuntu highly recommended. Also make sure to adjust tmpfs (/dev/shm) sizing based on memory size

./PROCEDURE: Once SETUP is fully complete, this brief guide will show you how to start fuzzing and more

If you like to test the MariaDB, MySQL or Percona server using the pquery framework instead, you can use this course:
(If you follow all steps in ./SETUP your server will already be setup for both the pquery and the Squirrel frameworks to run)
https://www.percona.com/blog/2015/03/17/free-mysql-qa-and-bash-linux-training-series/

If you get stuck, often the best way is to first view the script you were just using/running. There is a plethora of inline
documentation within scripts which may help one better understand what is happening or why something was done in a certain way

When you find a bug, please create a JIRA ticket in the MDEV project at https://jira.mariadb.org/ (account creation required)

Additionally, if you regularly test MariaDB, then please submit pull requests to include the UniqueID's for these bugs to:
https://github.com/mariadb-corporation/mariadb-qa/blob/master/known_bugs.strings 

And to include the required SQL for the bugs to mariadb-qa/BUGS:
https://github.com/mariadb-corporation/mariadb-qa/tree/master/BUGS

Note that doing so benefits the whole community of testers: any bug added to the UniqueID list will automatically filter
from all other testing instances where the filters are being used.

Feel free to contact https://github.com/mariadb-RoelVandePaar for more info, or feel free to create a issue ticket in 
https://github.com/mariadb-corporation/mariadb-qa if you find a bug, run into any issues, or have any questions!

Enjoy!