[erlang-questions] Fire and Forget

[Steven Livingstone]

Interesting - thanks Felix. The 50 machine cluster destruction is
quite an achievement :-) I'll take heed.

I could be hooking some devices in and so there could be quite a lot
of data as they scale. I am considering putting something like
RabbitMQ in the middle if i need to scale but starting with the basics
just now.

I am new to Erlang but the entire approach sits well with me so i want
to persist with it. The defaults are something i'll research - some
endpoint will be on Docker instances and i know there could be
additional things to figure out there too.

Thanks again,
steven

On Fri, Apr 1, 2016 at 5:54 PM, Felix Gallo <felixgallo@REDACTED> wrote:
> Also, as far as resources go - http clients by themselves are unlikely to
> tax a box at any rate you are likely to do. At high rates, you may encounter
> file handle or socket starvation on (at least) essentially all Linux boxes
> owing to pathologically poor defaults out of the box.  You may need to tweak
> sysctl values on those systems if you need more than a few thousand http
> requests per second.
>
> F.
>
> On Apr 1, 2016 9:44 AM, "Felix Gallo" <felixgallo@REDACTED> wrote:
>>
>> I accidentally once destroyed a 50 machine php http api cluster by
>> pointing a single four core erlang instance at it and making requests as
>> fast as I could go. You may wish to build rate limiting and error handling
>> in from the start instead of ignoring the response.
>>
>> F.
>>
>> On Apr 1, 2016 9:14 AM, "Steven Livingstone" <steven@REDACTED> wrote:
>>>
>>> Thanks Paul. It was more that as i would be just logging lots of data,
>>> if some gets lost it is ok, hence it doesn't need to wait on a 200.
>>>
>>> So, I guess it was how to optimise under those conditions, but perhaps
>>> doing an async post and later getting a 200, even if it is just a "all
>>> is ok" even if i don't intend to do anything beyond that may be a
>>> sensible approach anyway.
>>>
>>> Restart temporary under supervisor sounds good - i mainly don't want
>>> to block anything. Unfortunately I come with the burden of years of
>>> procedural programming so my mindset is still adjusting to the fact
>>> spinning up many processes is a good thing in erlang.
>>>
>>> It may be I should be passing the data using something other than an
>>> HTTP POST and more erlang-like but one step at a time :-)
>>>
>>> regards,
>>> steven
>>>
>>> On Fri, Apr 1, 2016 at 4:55 PM, Paul Peregud <paulperegud@REDACTED>
>>> wrote:
>>> > Low complexity solution: just spawn a process which will send request
>>> > and terminate.
>>> > More complex solution: spawn process under supervisor. Restart
>>> > strategy "temporary" is a good fit. Or "transient" if you want to
>>> > restart failed attempts.
>>> >
>>> > As for "response never happens". Do you mean that server does not
>>> > return anything ever? Not even HTTP 200 OK? Just holds up connection?
>>> >
>>> > On Fri, Apr 1, 2016 at 5:18 PM, Steven Livingstone <steven@REDACTED>
>>> > wrote:
>>> >> Hi all, Erlang newbie here. So, hello all :-) Now,
>>> >>
>>> >> I wish to do a fire and forget http post in Erlang.
>>> >>
>>> >> I looked at the httpc module which seems to allow synchronous and
>>> >> asynchronous calls but in the latter case i do not expect a return
>>> >> value so don't need a request id .... and not sure whether that
>>> >> behaviour influences what I want ... does it hold open a connection or
>>> >> use up resources on a response that is never going to happen etc.
>>> >>
>>> >> Does anyone know the best approach?
>>> >>
>>> >> many thanks,
>>> >> steven
>>> >> _______________________________________________
>>> >> erlang-questions mailing list
>>> >> erlang-questions@REDACTED
>>> >> http://erlang.org/mailman/listinfo/erlang-questions
>>> >
>>> >
>>> >
>>> > --
>>> > Best regards,
>>> > Paul Peregud
>>> > +48602112091
>>> _______________________________________________
>>> erlang-questions mailing list
>>> erlang-questions@REDACTED
>>> http://erlang.org/mailman/listinfo/erlang-questions