<div dir="ltr"><div><div><div>Thanks Fred.<br><br></div>I restart the node due to high memory usage, today everything looks fine. <br><br>Will update the data if it happen again.  <br><br></div>Thanks,<br></div>Linbo<br><div><div><br><br><br></div></div></div><div class="gmail_extra"><br><div class="gmail_quote">2015-12-08 3:30 GMT+08:00 Fred Hebert <span dir="ltr"><<a href="mailto:mononcqc@ferd.ca" target="_blank">mononcqc@ferd.ca</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">On 12/07, linbo liao wrote:<br>
</span><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">
Our Erlang server looks have a serious memory leak, the VM memory usage is<br>
low but top is high.<br>
<br></span>
*# Fragmentation*<span class=""><br>
<br>
I execute recon_alloc:fragmentation(current) and<br>
recon_alloc:fragmentation(max), find some allocator current usage is lower<br>
than max usage.<br>
<br></span>
*## Current usage*<span class=""><br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
     {{binary_alloc,0},<br>
      [{sbcs_usage,1.0},<br>
       {mbcs_usage,0.037804497612847224},<br>
       {sbcs_block_size,0},<br>
       {sbcs_carriers_size,0},<br>
       {mbcs_block_size,178384},<br>
       {mbcs_carriers_size,4718592}]},<br>
     {{binary_alloc,2},<br>
      [{sbcs_usage,2.0},<br>
       {mbcs_usage,0.05326200786389803},<br>
       {sbcs_block_size,0},<br>
       {sbcs_carriers_size,0},<br>
       {mbcs_block_size,4775112},<br>
       {mbcs_carriers_size,89653248}]},<br>
     {{binary_alloc,1},<br>
      [{sbcs_usage,2.0},<br>
       {mbcs_usage,0.0643930146188447},<br>
       {sbcs_block_size,0},<br>
       {sbcs_carriers_size,0},<br>
       {mbcs_block_size,4456384},<br>
       {mbcs_carriers_size,69206016}]},<br>
<br>
</blockquote></span></blockquote>
<br>
Yeah, those are very, very low usage values on mbcs (<5%). Can you give information such as the average block size (also in recon_alloc), how long the node has been running, and so on?<br>
<br>
All values for current and max are useful for these. Also if you could provide your allocator strategy that could be useful.<br>
</blockquote></div><br></div>