[IPython-User] ipcluster runs out of memory and can't purge results
Sun Dec 16 05:39:09 CST 2012
On 15/12/2012 14:22, Johann Rohwer wrote:
> On 15/12/2012 01:11, Fernando Perez wrote:
>> On Fri, Dec 14, 2012 at 8:25 AM, Robert Nishihara
>> <email@example.com> wrote:
>>> Not sure if this will solve your problem, but I think I've run into a
>>> similar problem, and I've had some success with this.
>> Please keep us posted on how this goes, so we can offer a more robust
>> and easy to use solution out of the box: if Robert's solution works
>> for you we can wrap it into a user-facing utility.
>> Also, does running %reset help any?
> %reset by itself seems to be doing the trick for me without
> needing all the commands in Robert's solution!
I posted too early and carelessly. I had looked at the wrong column in
'top'. %reset does *not* free the memory on its own, and this does make
sense as I had tried myself to manually delete the rc and lv objects from
the namespace, to no avail (see earlier post in thread).
The good news is that Robert's clear_cache function does indeed work, and
that I can now run my simulation to completion without the total memory used
increasing when running successive chunks! So I guess it would be a great
idea to wrap this function into a user-facing utility.
On an aside, can anyone explain to me why this is necessary, and why
deleting the rc and lv objects from the namespace and calling gc.collect()
is insufficient? I am really not getting this...
More information about the IPython-User