Hi all,
I've just changed my django app to use multiprocessing and Pool. On my dev machine (OSX) this work fine. When I deploy to my Linode however, the underlying os.fork() method throws:

OSError: [Errno 12] Cannot allocate memory

Anyone know much about this? My Linode is configured with Ubuntu 11.04. It has 512mb RAM, but this doesn't seem to get used according to top. However, the output from free -m is:

total used free shared buffers cached
Mem: 497 488 9 0 1 42
-/+ buffers/cache: 443 53
Swap: 255 255 0

Which looks rather bad. Anyone got any suggestions? I just don't think all my memory is gone! According to top Apache is holding 17.5% and a few other processes have around 10% in total.

Ta.

Here's some more info. I rebooted the server and reloaded the page that does the processing a few times, logging the memory availability after each reload. It seems that somewhere in apache/python/django it is not freeing up the memory after each load:

total used free shared buffers cached
Mem: 497 213 283 0 66 51
-/+ buffers/cache: 96 401
Swap: 255 0 255

total used free shared buffers cached
Mem: 497 348 149 0 66 82
-/+ buffers/cache: 199 298
Swap: 255 0 255

total used free shared buffers cached
Mem: 497 462 35 0 66 84
-/+ buffers/cache: 310 187
Swap: 255 0 255

total used free shared buffers cached
Mem: 497 486 10 0 11 53
-/+ buffers/cache: 421 76
Swap: 255 6 249

total used free shared buffers cached
Mem: 497 490 7 0 0 19
-/+ buffers/cache: 470 27
Swap: 255 155 100

total used free shared buffers cached
Mem: 497 474 22 0 1 25
-/+ buffers/cache: 448 48
Swap: 255 254 1

I saw this behavior using multiprocessing pools on my Django website http://zunzun.com, and I kill the process after each use and mod_wsgi respawns a new process for me automatically. That worked, so I stopped investigating.

James