[erlang-questions] poolboy new_worker failed cause poolboy crashed

[linbo liao]

I find the issue, refre to poolboy checkin (
https://github.com/devinus/poolboy/blob/master/src/poolboy.erl#L309), when
checkin process and overflow >0, it will close process to make sure pool
size return back, won't overflow.  It will cause process be closed (maybe
old process, maybe new process).

Thanks,
Linbo

2015-04-16 9:19 GMT+08:00 linbo liao <llbgurs@REDACTED>:

> Hi José,
>
> I add a proxy process to handle the issue (
> https://github.com/yunba/eredis_pool/blob/master/src/eredis_proxy.erl),
> but wired if connect redis timeout, proxy process sometimes will be
> shutdown although I set trap_exit = true.
>
> I test it , find all connections timeout when multi client call connnect
> simulataneous, after eredis_client exit info received, eredis_proxy will be
> shutdown. I have totally no idea about it.
>
> Did I miss something?
>
> Thanks,
> Linbo
>
> 2015-04-14 19:01 GMT+08:00 linbo liao <llbgurs@REDACTED>:
>
>> Thanks, looks a nice way to solve the issue, I will create a proxy worker
>> to have a try.
>>
>> 2015-04-14 16:49 GMT+08:00 José Valim <jose.valim@REDACTED>:
>>
>>> You don't need to modify eredis per se. You can create a worker that is
>>> between poolboy and eredis that tracks the connected/disconnected semantics
>>> as Devin described. Unfortunately I don't know of any sample Erlang code
>>> but, if you don't mind, here is a similar solution using redo written in
>>> Elixir:
>>>
>>>
>>> https://github.com/phoenixframework/phoenix/blob/master/lib/phoenix/pubsub/redis_conn.ex
>>>
>>> To use it, you checkout the worker above from the pool, ask the redis
>>> connection to the worker and do the operations you want to in the
>>> connection as usual.
>>>
>>> As a note, I did find it a bit weird that eredis handles disconnects
>>> automatically but it does not handle the case where it can't connect to
>>> Redis on start_link. It would be handful if they supported lazy connections
>>> (i.e. don't connect on start_link but on first command). They already have
>>> all the code to handle this, it is just a matter of supporting it on
>>> start_link too.
>>>
>>>
>>>
>>> *José Valim*
>>> www.plataformatec.com.br
>>> Skype: jv.ptec
>>> Founder and Lead Developer
>>>
>>> On Tue, Apr 14, 2015 at 10:26 AM, linbo liao <llbgurs@REDACTED> wrote:
>>>
>>>> Hi,
>>>>
>>>> I open an issue for poolboy (
>>>> https://github.com/devinus/poolboy/issues/72), but looks author
>>>> doesn't think this is an issue.
>>>>
>>>> Looks like to solve this issue have two ways:
>>>>
>>>> 1. Modify eredis client,  start_link still return ok when connect
>>>> failed, and when use it should check eredis socket alive or not
>>>>
>>>> 2. Modify poolboy library, protected all logic when call new_worker
>>>> failed.
>>>>
>>>> But it require change ereids or poolboy code, I was wondering to know
>>>> anyone hit this issue, how to resolve it?
>>>>
>>>> Thanks,
>>>> Linbo
>>>>
>>>> _______________________________________________
>>>> erlang-questions mailing list
>>>> erlang-questions@REDACTED
>>>> http://erlang.org/mailman/listinfo/erlang-questions
>>>>
>>>>
>>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20150416/c3b93a08/attachment.htm>