This module is an NFS client for python. This project is hosted at https://github.com/sahlberg/libnfs-python
This module depends on the libnfs library. You must first install libnfs before you can use this module. Libnfs can be found at https://github.com/sahlberg/libnfs
The two main classes in this module are NFS and NFSFH. NFS represents an NFS context that describes a mount point. NFSFH is the main class that implements a file-like object for NFS files.
This module is distributed under LGPL version 2.1. Please see COPYING for the full text of this license.
In the context-full mode you will first need to mount an NFS share and create an NFS context before you can access any files. This NFS context is then used for all future access to files in that mount point.
The benefit with this mode is that since performing the actual NFS mount is very expensive you only need to perform it once, when creating the NFS object and can reuse the mount for any future files you want to access.
In this mode absolute paths are relative to the mount point for the NFS object.
import libnfs nfs = libnfs.NFS('nfs://127.0.0.1/data/tmp/') a = nfs.open('/foo-test', mode='w+') a.write("Test string") a.close() print nfs.open('/foo-test', mode='r').read()
This module also provides a mode where you can directly open and access a file-like object without first having to instantiate a NFS mount. In this case the file-like object will internally create an NFS mount for the duration of the lifespan of the file-like object.
In this mode you invoke the libnfs.open() function with an NFS url that points to a file :
import libnfs a = libnfs.open('nfs://127.0.0.1/data/tmp/foo-test', mode='w+') a.write("Test string") a.close() print libnfs.open('nfs://127.0.0.1/data/tmp/foo-test', mode='r').read()
In the example above we create two instances of a file-like object. First we create and write to the file and second time we read the file back and print it. This is however both expensive and potentially poor performing. The reason is that when we use context-free file-like objects we have to create an NFS mount internally for each object and this is expensive.
Basically, when creating an NFS mount, the NFS protocol requires that we perform the following operations/roundtrips :
1, Establish TCP connection to portmapper (TCP port 111) on the server. 2, Send a Portmapper NULL command to verify we are connected to a genuine Portmapper and that it is alive. 3, Send a Portmapper GetPort command to fetch which port the MOUNT daemon is running on. Close the TCP connection. 4, Establish a TCP connection to the MOUNT daemon 5, Send a MOUNT NULL command to verify the MOUNT daemon is alive. 6, Send a MOUNT MNT command to fetch the NFS file handle for the mount point. Close the TCP connection. 7, Establish TCP connection to portmapper (TCP port 111) on the server. 8, Send a Portmapper NULL command to verify we are connected to a genuine Portmapper and that it is alive. 9, Send a Portmapper GetPort command to fetch which port the DAEMON daemon is running on. Close the TCP connection. 10, Establish a TCP connection to the NFS daemon 11, Send a NFS NULL command to verify the NFS daemon is alive. 12, Send a NFS FSINFO command to fetch some basic information about the filesystem we just mounted. 13, Send a NFS GETATTR command to fetch some information about the root directory for this mount.
When using context-free handles we have to perform these steps every time a new file-like object is created making it very expensive.
The benefit with context-free handles is that they are very convenient to use. Just pass an NFS url to the libnfs.open() function and you have a file-like object. The drawback is that every context-free handle have to perform their own NFS mount internally. If you plan to access multiple/many files you should consider using context-full handles instead.
A libnfs-python mailing list is available at http://groups.google.com/group/libnfs-python Announcements of new versions of libnfs-python will be posted to this list.