wizhacks / chorddfs

ChordDFS

Distributed File System implementation using Chord algorithm

Requirements

• Mininet 2.1.0
• MiniNExT
• Python 2 or 3

Instructions

1. Run sudo python start.py --num_nodes where num_nodes is the number of nodes in your topology that you want to start.
2. Start Client protocol on node 2 by running node# python Chord.py n2 \n2; ie, n2 python Client.py n2 \n2.

Note that currently n1 is the tracker node and n2 is set up to be the client node by default.

Examples

n1 python Chord.py n1 \n1 &					# n1 run server in background
n2 python Client.py n2 \n2					# n2 run client with stdin i/o
n3 python Client.py n3 \n3 script.txt 		# n3 run client with script, no i/o

chorddfs's Issues

chord.py

Client program crashes mininet when client crashes or call exit

1. start client program ie n1 python Client.py n1 \n1
2. type exit into stdin console
3. the client program should exit with this message on the screen: "bash: error reading input file: Resource temporarily available"
4. n1 any_cmd
5. mininet crashes with OSError: Broken Pipe

client.py

config

Flushing data error

When log files reach ~70 bytes, there seems to be an issue with flushing the buffer. Since this throws an error, the msg is not written to log either. We can try switching order or catching error?

<172.1.1.1>: msg type:RETURN_PREDECESSOR rcvd from 172.1.1.3: msg:{u'client_ip': None, u'target': None, u'msg_type': u'RETmininext> n4 ls
p': None, 'finger': None, 'key': None}
Traceback (most recent call last):
  File "Chord.py", line 453, in <module>
    ctrlMsgReceived()
  File "Chord.py", line 93, in ctrlMsgReceived
    myLogger.mnPrint("msg type:{0} rcvd from {1}: msg:{2}".format(msg_type, addr[0], msg))
  File "/home/mininet/ChordDFS/ReadLog.py", line 23, in mnPrint
    sys.stdout.flush() # need to flush output, else never show up
IOError: [Errno 4] Interrupted system call

Report Errors to Client

If a file doesn't exist on the network for a get request, we should return an error to the Client. Same for other issues.

topo.py

tracker, client, chord nodes, fabric switch

Can't have more than 10 nodes connected to fabric switch, times out and crashes