<div dir="ltr">Thank you Anton,<div>I do not have such system yet in place, as this isn't a production environment (yet).</div><div><br></div><div>Will definitely add one asap.</div><div><br></div><div>Thanks,</div><div>r.</div><div><br></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Jan 22, 2015 at 7:41 PM, Anton Lebedevich <span dir="ltr"><<a href="mailto:mabrek@gmail.com" target="_blank">mabrek@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="HOEnZb"><div class="h5">Hello.<br>
<br>
Do you have any external monitoring system in place (like statsderl or<br>
exometer + graphite) that watches erlang memory consumption (process,<br>
binary, ... total)? What is the long-term growth trend? Is it linear<br>
or similar to capacitor charge (like<br>
<a href="http://hades.mech.northwestern.edu/index.php/File:RC_charge_voltage.gif" target="_blank">http://hades.mech.northwestern.edu/index.php/File:RC_charge_voltage.gif</a>)<br>
<br>
>From the data provided it looks like some subpopulation of processes<br>
is growing their heaps. It might happen for mid- or small-sized<br>
processes since top processes always have the same size.<br>
Can you measure [memory, message_queue_len, total_heap_size,<br>
heap_size, stack_size] for all processes several times (with the same<br>
interval as you measured total memory growth) and compare histograms<br>
of their memory consumption? Or upload that data somewhere and I'll<br>
play with it in R to find what's changing inside.<br>
<br>
<br>
Regards,<br>
Anton Lebedevich.<br>
<br>
</div></div>PS forgot to reply to the list<br>
</blockquote></div><br></div>