A remotely controlled fault injection file system.
$ pyenv virtualenv 3.8.4 charybdisfs
$ sudo apt install build-essential pkg-config libfuse3-dev
$ sudo yum install fuse3 fuse3-devel
$ pyenv local charybdisfs
$ pip install -r requirements.txt
$ docker build -t charybdisfs .
All CharybdisFS mounts will be available inside Docker container only.
$ docker run -it --rm --device /dev/fuse --privileged -v $(pwd):/src charybdisfs bash
Whole host filesystem will be available at /docker_host_root and if you mount CharybdisFS at any point under this directory then it'll be available on the host system too.
$ docker run -it --rm --device /dev/fuse --privileged \
-v $(pwd):/src -v /:/docker_host_root:rw,rshared charybdisfs bash
The idea behind is to create a bind mount to have access to original files and then mount CharybdisFS to the original path. Note that bind mounts can be done only on the same filesystem.
$ mkdir /path/to/.shadow_source_dir
$ mount --bind /path/to/source_dir /path/to/.shadow_source_dir
$ ./charybdisfs /path/to/.shadow_source_dir /path/to/source_dir
Unfortunately, you can't use this trick with Docker container mount propogation together.
Import all libs
from client import CharybdisFsClient
from core.faults import LatencyFault, ErrorFault, SysCall
import errno
Connect to the rest server
fs_client = CharybdisFsClient('127.0.0.1', 8080)
Create fault object and add it - succeeded
latency_fault = LatencyFault(sys_call=SysCall.WRITE, probability=100, delay=1000)
fault_id, resp = fs_client.add_fault(latency_fault)
Expected result:
fault_id
'3af4e469-5e36-4d6c-99a1-1919944e6419'
resp
<Response [200]>
Create fault object and add it - failed
error_fault = ErrorFault(sys_call=SysCall.WRITE, probability=100, error_no=errno.EADV)
f_id2, resp2 = fs_client.add_fault(error_fault)
Expected result:
f_id2
''
resp2
<Response [500]>
Remove fault
l = fs_client.remove_fault('3af4e469-5e36-4d6c-99a1-1919944e6419')
Expected result:
l
<Response [200]>
l.text
'{"fault_id": "3af4e469-5e36-4d6c-99a1-1919944e6419"}'