DecaFS

Distributed educational component adaptable File System

Starting DecaFS

First, install the executables.

1. Copy the DecaFS source to the target system (e.g. one of your Raspberry Pis).
2. Run make from the root of the source tree. This might take a while, but should complete successfully.
3. Copy bin/decafs_barista to the home directory on one node, and bin/decafs_espresso to the home directory on each of the others. With appropriate modifications, scripts/distribute_exe.sh will do this.

Then, use the startup script either directly from one of your nodes or from a separate system.

1. Make a copy of scripts/sample_config.json as e.g. scripts/config.json, changing the IP addresses to point to the nodes that have the appropriate executables.
2. Run python scripts/startup.py --config scripts/config.json. This should print messages indicating that each of nodes 1-4 has started successfully.

Alternately, you can use tmuxinator to launch DecaFS. Running tmuxinator decafs_local from the root of the source tree will launch a local instance of DecaFS. The configuration file scripts/decafs.yml can be placed in ~/.tmuxinator/ and modified as appropriate to launch DecaFS on your cluster.

Troubleshooting DecaFS

If DecaFS doesn't start correctly, this may be due to

1. Leftover TCP connections from a recent run. The barista will print "Address already in use", espressos will print "Connection refused", and (on Linux) netstat -atn|grep 3899 will show sockets in TIME_WAIT state. Wait a minute or two and try again.
2. Corrupted persistent metadata. The barista will immediately crash with a segmentation fault. Delete decafs_metadata* files on the barista node (and optionally the espresso nodes, for a more complete reset) and try again.

Testing DecaFS

Once DecaFS is running, you can validate using a test client. From another terminal:

1. Go to src/app/decafs_test_client
2. Change the IP address in main.cpp to point to the barista node
3. make
4. Run ./nettest