I’ve been toying with an idea of exposing statistics for a Python application via shared memory to keep the performance impact on the application as low as possible. The goal being an application could passively expose a number of metrics that could either be periodically polled via munin/Icinga/etc plugins or interactive tools when diagnosing issues on a system.
Usually in the UNIX world you have 2 ways of accessing/manipulating data: memory addresses or streams (files). Manipulating data via memory addresses means pointers, offsets, malloc/free, etc. Stream interfaces manipulate data via read/write/seek system calls for files and send/recv/etc for sockets.
mmap gives you both interfaces. A memory mapped file can be manipulated via read/write/seek or by directly accessing its mapped memory region. The advantage of the latter is that this memory region is in userspace — meaning you can manipulate a file without incurring the overhead of write system calls for every manipulation.
Anyway, enough exposition, let’s see some code. (Despite mmap’s nice featureset, I’m only using it as a simple memory sharing mechanism anyway.) The following code shares a tiny bit of data between 2 Python processes using the excellent mmap module in the stdlib.
a.py writes to the memory mapped region, and
b.py reads the data out. ctypes allows for an easy way to create values in a memory mapped region and manipulate them like “normal” Python objects.
These code samples were written using Python 2.7 on Linux. They should work fine on any POSIX system, but Windows users will have to change the mmap calls to match the Windows API.
#!/usr/bin/env python import ctypes import mmap import os import struct def main(): # Create new empty file to back memory map on disk fd = os.open('/tmp/mmaptest', os.O_CREAT | os.O_TRUNC | os.O_RDWR) # Zero out the file to insure it's the right size assert os.write(fd, '\x00' * mmap.PAGESIZE) == mmap.PAGESIZE # Create the mmap instace with the following params: # fd: File descriptor which backs the mapping or -1 for anonymous mapping # length: Must in multiples of PAGESIZE (usually 4 KB) # flags: MAP_SHARED means other processes can share this mmap # prot: PROT_WRITE means this process can write to this mmap buf = mmap.mmap(fd, mmap.PAGESIZE, mmap.MAP_SHARED, mmap.PROT_WRITE) # Now create an int in the memory mapping i = ctypes.c_int.from_buffer(buf) # Set a value i.value = 10 # And manipulate it for kicks i.value += 1 assert i.value == 11 # Before we create a new value, we need to find the offset of the next free # memory address within the mmap offset = struct.calcsize(i._type_) # The offset should be uninitialized ('\x00') assert buf[offset] == '\x00' # Now ceate a string containing 'foo' by first creating a c_char array s_type = ctypes.c_char * len('foo') # Now create the ctypes instance s = s_type.from_buffer(buf, offset) # And finally set it s.raw = 'foo' print 'First 10 bytes of memory mapping: %r' % buf[:10] raw_input('Now run b.py and press ENTER') print print 'Changing i' i.value *= i.value print 'Changing s' s.raw = 'bar' new_i = raw_input('Enter a new value for i: ') i.value = int(new_i) if __name__ == '__main__': main()
import mmap import os import struct import time def main(): # Open the file for reading fd = os.open('/tmp/mmaptest', os.O_RDONLY) # Memory map the file buf = mmap.mmap(fd, mmap.PAGESIZE, mmap.MAP_SHARED, mmap.PROT_READ) i = None s = None while 1: new_i, = struct.unpack('i', buf[:4]) new_s, = struct.unpack('3s', buf[4:7]) if i != new_i or s != new_s: print 'i: %s => %d' % (i, new_i) print 's: %s => %s' % (s, new_s) print 'Press Ctrl-C to exit' i = new_i s = new_s time.sleep(1) if __name__ == '__main__': main()
(Note that I cruelly don’t clean up /tmp/mmaptest after the scripts finished. Consider it a 4KB tax for anyone who runs arbitrary code they found on the Internet without reading it first.)