pyfs_stress

Multi-client file system load and stress testing tool based on ZeroMQ.

It could be useful for anyone, who is developing new file system or testing an existing one.

Use cases:

1. Load & Stress - Imagine 60 clients, each one running 16 io processes vs your file server :)
2. Race Conditions - Client x trying to access file which is removed by client y, while client z reading it
3. Data corruptions detection - Controller always "knows" the current status of all files and directories

Features:

• Multi-client load & stress - tested on 60 clients simultaneously ( and it's still not the limit :))
• Data integrity & corruptions monitoring - able to detect lost files or file chunks, containing wrong data
• NFS and SMB mounts are supported
• Following file system operations supported: mkdir, list, remove, touch, stat, read, direct read, rename, move, write, truncate
• Random and sequential reads and writes supported
• NFS advisory file locking
• Centralised logging (PUB - SUB) for all clients
• Configurable workloads (JSON)
• Flexible configuration (JSON)

Linux distributions tested: CentOS 6, CentOS 7, Debian Jessy, Ubuntu 14.04

System Requirements:

• File Server under test should support nfs or smb protocol
• At least 2 physical machines or VMs for Sever (controller) and one client
• At least 8GB RAM for server (controller) machine
• At least 4GB RAM for client machine

Required Python packages:

• pyzmq
• paramiko
• treelib
• argparse
• xxHash

Usage:

]# ./fileops_server.py --help
usage: fileops_server.py [-h] [-c CLIENTS [CLIENTS ...]] [-e EXPORT]
                         [--start_vip START_VIP] [--end_vip END_VIP]
                         [--tenants] [-m {nfs3,nfs4,nfs4.1,smb1,smb2,smb3}]
                         cluster

pyFstress Server runner

positional arguments:
  cluster               File server name or IP

optional arguments:
  -h, --help            show this help message and exit
  -c CLIENTS [CLIENTS ...], --clients CLIENTS [CLIENTS ...]
                        Space separated list of clients
  -e EXPORT, --export EXPORT
                        NFS export name
  --start_vip START_VIP
                        Start VIP address range
  --end_vip END_VIP     End VIP address range
  --tenants             Enable MultiTenancy
  -m {nfs3,nfs4,nfs4.1,smb1,smb2,smb3}, --mtype {nfs3,nfs4,nfs4.1,smb1,smb2,smb3}
                        Mount type

Feature plans:

• Implement windows clients support
• Improve data integrity check
• Performance stats
• Additional workloads
• Performance optimisation

Notes:

This project is still under active development so feel free to test it, open issues or make pull requests.