cfs (cloud file system) is a low-level filesystem interface for building the cloud. cfs is not a distributed file system. It is designed to be the lowest building block of cloud infrastructure that is available on all machines. The design goal of cfs is reliability, simplicity and usability.

Current status is pre-alpha

TODO:

1. basic file system interface (write, read, remove, mkdir, rename)
2. copy from another cfs server
3. bit-matrix based data reconstruction
4. expose basic run time metrics
5. basic resource enforcement
6. container deployment

alpha

TODO:

1. profiling and benchmark
2. per disk metadata store
3. simple ACL support
4. user level buffer
5. tons of tests

We will BREAK everything (data format, API, command line tool, etc.) before 1.0.

cfs depends on GF-complete and Jerasure for erasure coding.

To install GF-complete, see https://github.com/c-fs/gf-complete

To install Jerasue, see https://github.com/c-fs/Jerasure/

TODO: write build/MAKE file

cd server

go build

./server

cfs listens on 15524 by default. The root path of cfs is the current directory by default.

cfs creates a /cfs0 disk with root path /cfs0000 by default.

TODO: make all these configurable.

cd cfsctl

go build

./cfsctl -action write -name "cfs0/foo" -data "bar"
2015/05/24 11:16:48 3 bytes written to foo at offset 0

./cfsctl -action read -name "cfs0/foo" -length 100
2015/05/24 11:17:34 bar

echo "corrupt" -> ./server/cfs0000/foo

Try to read out the file again

./cfsctl -action read -name "cfs0/foo" -length 100
2015/05/24 11:18:56

Nothing is read out.

We can see the log output from the server side

2015/05/24 11:18:56 server: read error disk: not a valid CRC