script uses up all memory

[Larry Martell]

On Wed, Mar 5, 2014 at 7:33 PM, Chris Angelico <rosuav at gmail.com> wrote:
> On Thu, Mar 6, 2014 at 11:20 AM, Larry Martell <larry.martell at gmail.com> wrote:
>> On Wed, Mar 5, 2014 at 5:39 PM, Chris Angelico <rosuav at gmail.com> wrote:
>>> On Thu, Mar 6, 2014 at 9:27 AM, Larry Martell <larry.martell at gmail.com> wrote:
>>>> I have a script that forks off other processes and attempts to manage
>>>> them. Here is a stripped down version of the script:
>>>>
>>>>         self.sleepTime = 300
>>>
>>> That's not a stand-alone script.
>>
>> No, that is just the part that does the work (inside the 'while
>> true'). I'll try and post a standalone script tomorrow.
>>
>>> What environment is it running in?
>>
>> CentOS 6.4
>
> That's not the whole environment, though. There's a mention of Django
> - does this run inside some framework?

The system this is part of uses Django, and this script makes use of
the django ORM, but it doesn't do any web stuff itself. It just kicks
off another script once for each tool found in the database, and
ensure that there's just one script pre tool running at a time, and
that no single script runs too long. The django part can easily be
removed.

>>> Can you reproduce the problem outside of that environment?
>>
>> I will try that tomorrow.
>
> Running as a stand-alone script, still under CentOS, would be what I
> mean by "outside of that environment". I'm talking about making
> something that can be saved to my drive and executed, perhaps with a
> stubby subprocess script (eg "import time; time.sleep(86400)").

Yes, I understand what you mean.

>
>>> Also: Can you simply use multiprocessing rather than going through all
>>> the effort of subprocess.Popen?
>>
>> Perhaps. I didn't write this. A client gave it to me and said 'figure
>> out why it uses up all the memory and hangs.' I've messed around with
>> for days and cannot see anything that would consume so much memory.
>
> Ah. Yeah, that would be a fun little job to play with. My random
> thought: Do the subprocesses produce spammy log output? If so, the
> monitor might be collecting it all and holding it in memory in case
> you want it (not knowing that you don't). The default should be to
> leave them connected to your process's stdio streams, though, so that
> shouldn't be the issue.

Ohh, that's a good random thought. I'll try that tomorrow and see if
that's the issue.