Problem with boss_news and boss_db

[Jose Luis Gordo Romero]

Hi,

I'm having problems with boss_news, I declared in the init the following:

init() ->

    {ok, Acty} = boss_news:watch("modelxs",

        fun (created, X) ->

            {ok, ActyC} = activities:add(X)

        end),

    {ok, [Acty]}.

activities:add(X) only adds a new record to a different model.

When a record is inserted in modelx, the run_after_hooks timeouts:

** exception exit: {timeout,{gen_server,call,

                                        [{global,boss_news},

                                         {created,"modelx-43",

                                                  [{id,"modelx-43"},

                                                   .....

                                                   {created_time,{1328,815623,280614}},

                                                   {updated_time,{1328,815623,280614}}]}]}}

     in function  gen_server:call/2 (gen_server.erl, line 180)

     in call from boss_record_lib:run_after_hooks/3 (src/boss/boss_record_lib.erl, line 11)

     in call from boss_db:save_record/1 (src/boss/boss_db.erl, line 195)

Only happens when creates a new boss_db record inside the boss_new:wath/2 fun, I'm in a multiapp setup (but started-dev from the app containing the init watches) and working with one node (dev).

The funny thing is that the new record is created, but the timeout+restart makes this ugly. 

Any thoughts?

Thanks!

--

Jose Luis Gordo Romero

@jgordor - http://www.freemindsystems.com

[Evan Miller]

I think the problem is that the boss_news process is executing the callback, but the callback then sends a message to boss_news => deadlock. This is a design flaw. We need worker processes to execute the callbacks instead

Sent from my iPhone

[Jose Luis Gordo Romero]

Thanks Evan,

A good moment to start using https://github.com/devinus/poolboy ?

--

Jose Luis Gordo Romero

@jgordor - http://www.freemindsystems.com

2012/2/9 Evan Miller <emmi...@gmail.com>

[douglas]

In the spirit of ChicagoBoss , why not call the ORM dempster_diver

[Evan Miller]

On Fri, Feb 10, 2012 at 1:21 AM, Jose Luis Gordo Romero
<jgo...@gmail.com> wrote:
> Thanks Evan,
>
> A good moment to start using https://github.com/devinus/poolboy ?
>

Not necessarily. We might want an unlimited number of workers created
on-demand rather than a fixed pool size.

I'll try to look into fixing the process model of all of CB's services
this weekend.

Evan

--
Evan Miller
http://www.evanmiller.org/

[Jose Luis Gordo Romero]

This sounds amazing!

--

Jose Luis Gordo Romero

@jgordor - http://www.freemindsystems.com

2012/2/10 Evan Miller <emmi...@gmail.com>

[douglas]

mongos

For sharded environments, each mongos process can perform any number of operations concurrently. This results in downstream operations to mongod instances. Execution of operations at each mongod is independent; that is, one mongod does not block another.

[Evan Miller]

Try this

https://github.com/evanmiller/boss_db/commit/8076dda5a23bb4e174e6226087b90fe8ceee2c2b

There are also some UNTESTED poolboy commits in there, but I think the
above patch will solve the BossNews deadlock you encountered.

[Jose Luis Gordo Romero]

Hi,

I'm not tested all the app, but I can confirm that the commit solves the deadlock.

Terrific!

--

Jose Luis Gordo Romero

@jgordor - http://www.freemindsystems.com

2012/2/11 Evan Miller <emmi...@gmail.com>