[erlang-questions] Possilbe Erlang memory fragmentation

[linbo liao]

Thanks Fred.

I restart the node due to high memory usage, today everything looks fine.

Will update the data if it happen again.

Thanks,
Linbo




2015-12-08 3:30 GMT+08:00 Fred Hebert <mononcqc@REDACTED>:

> On 12/07, linbo liao wrote:
>
>> Our Erlang server looks have a serious memory leak, the VM memory usage is
>> low but top is high.
>>
>> *# Fragmentation*
>>
>> I execute recon_alloc:fragmentation(current) and
>> recon_alloc:fragmentation(max), find some allocator current usage is lower
>> than max usage.
>>
>> *## Current usage*
>>
>>      {{binary_alloc,0},
>>>       [{sbcs_usage,1.0},
>>>        {mbcs_usage,0.037804497612847224},
>>>        {sbcs_block_size,0},
>>>        {sbcs_carriers_size,0},
>>>        {mbcs_block_size,178384},
>>>        {mbcs_carriers_size,4718592}]},
>>>      {{binary_alloc,2},
>>>       [{sbcs_usage,2.0},
>>>        {mbcs_usage,0.05326200786389803},
>>>        {sbcs_block_size,0},
>>>        {sbcs_carriers_size,0},
>>>        {mbcs_block_size,4775112},
>>>        {mbcs_carriers_size,89653248}]},
>>>      {{binary_alloc,1},
>>>       [{sbcs_usage,2.0},
>>>        {mbcs_usage,0.0643930146188447},
>>>        {sbcs_block_size,0},
>>>        {sbcs_carriers_size,0},
>>>        {mbcs_block_size,4456384},
>>>        {mbcs_carriers_size,69206016}]},
>>>
>>>
> 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?
>
> All values for current and max are useful for these. Also if you could
> provide your allocator strategy that could be useful.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20151208/a00e9eca/attachment.htm>