One of the things that attracted me to Erlang in the first place is the Actor model; the idea that different processes run concurrently and interact via asynchronous messaging.
I'm just starting to get my teeth into OTP and in particular looking at gen_server. All the examples I've seen - and granted they are tutorial type examples - use handle_call() rather than handle_cast() to implement module behaviour.
I find that a little confusing. As far as I can tell, handle_call is a synchronous operation: the caller is blocked until the callee completes and returns. Which seems to run counter to the async message passing philosophy.
I'm about to start a new OTP application. This seems like a fundamental architectural decision so I want to be sure I understand before embarking.
My questions are:
In real practice do people tend to use handle_call rather than handle_cast?
If so, what's the scalability impact when multiple clients can call the same process/module?
Depends on your situation.
If you want to get a result, handle_call is really common. If you're not interested in the result of the call, use handle_cast. When handle_call is used, the caller will block, yes. This is most of time okay. Let's take a look at an example.
If you have a web server, that returns contents of files to clients, you'll be able to handle multiple clients. Each client have to wait for the contents of files to be read, so using handle_call in such a scenario would be perfectly fine (stupid example aside).
When you really need the behavior of sending a request, doing some other processing and then getting the reply later, typically two calls are used (for example, one cast and the one call to get the result) or normal message passing. But this is a fairly rare case.
Using handle_call will block the process for the duration of the call. This will lead to clients queuing up to get their replies and thus the whole thing will run in sequence.
If you want parallel code, you have to write parallel code. The only way to do that is to run multiple processes.
So, to summarize:
Using handle_call will block the caller and occupy the process called for the duration of the call.
If you want parallel activities to go on, you have to parallelize. The only way to do that is by starting more processes, and suddenly call vs cast is not such a big issue any more (in fact, it's more comfortable with call).
Adam's answer is great, but I have one point to add
Using handle_call will block the process for the duration of the call.
This is always true for the client who made the handle_call call. This took me a while to wrap my head around but this doesn't necessarily mean the gen_server also has to block when answering the handle_call.
In my case, I encountered this when I created a database handling gen_server and deliberately wrote a query that executed SELECT pg_sleep(10), which is PostgreSQL-speak for "sleep for 10 seconds", and was my way of testing for very expensive queries. My challenge: I don't want the database gen_server to sit there waiting for the database to finish!
My solution was to use gen_server:reply/2:
This function can be used by a gen_server to explicitly send a reply to a client that called call/2,3 or multi_call/2,3,4, when the reply cannot be defined in the return value of Module:handle_call/3.
In code:
-module(database_server).
-behaviour(gen_server).
-define(DB_TIMEOUT, 30000).
<snip>
get_very_expensive_document(DocumentId) ->
gen_server:call(?MODULE, {get_very_expensive_document, DocumentId}, ?DB_TIMEOUT).
<snip>
handle_call({get_very_expensive_document, DocumentId}, From, State) ->
%% Spawn a new process to perform the query. Give it From,
%% which is the PID of the caller.
proc_lib:spawn_link(?MODULE, query_get_very_expensive_document, [From, DocumentId]),
%% This gen_server process couldn't care less about the query
%% any more! It's up to the spawned process now.
{noreply, State};
<snip>
query_get_very_expensive_document(From, DocumentId) ->
%% Reference: http://www.erlang.org/doc/man/proc_lib.html#init_ack-1
proc_lib:init_ack(ok),
Result = query(pgsql_pool, "SELECT pg_sleep(10);", []),
gen_server:reply(From, {return_query, ok, Result}).
IMO, in concurrent world handle_call is generally a bad idea. Say we have process A (gen_server) receiving some event (user pressed a button), and then casting message to process B (gen_server) requesting heavy processing of this pressed button. Process B can spawn sub-process C, which in turn cast message back to A when ready (of to B which cast message to A then). During processing time both A and B are ready to accept new requests. When A receives cast message from C (or B) it e.g. displays result to the user. Of course, it is possible that second button will be processed before first, so A should probably accumulate results in proper order. Blocking A and B through handle_call will make this system single-threaded (though will solve ordering problem)
In fact, spawning C is similar to handle_call, the difference is that C is highly specialized, process just "one message" and exits after that. B is supposed to have other functionality (e.g. limit number of workers, control timeouts), otherwise C could be spawned from A.
Edit: C is asynchronous also, so spawning C it is not similar to handle_call (B is not blocked).
There are two ways to go with this. One is to change to using an event management approach. The one I am using is to use cast as shown...
submit(ResourceId,Query) ->
%%
%% non blocking query submission
%%
Ref = make_ref(),
From = {self(),Ref},
gen_server:cast(ResourceId,{submit,From,Query}),
{ok,Ref}.
And the cast/submit code is...
handle_cast({submit,{Pid,Ref},Query},State) ->
Result = process_query(Query,State),
gen_server:cast(Pid,{query_result,Ref,Result});
The reference is used to track the query asynchronously.
Related
To process HTTP requests, we have to make blocking calls (e.g. JDBC calls) as part of a Mono/Flux-based process. Our current plan looks something like this:
// I renamed getSomething to processJaxrsHttpRequest
CompletionStage<String> processJaxrsHttpRequest(String input) {
return Mono.just(input)
.map(in -> process(in))
.flatMap(str -> Mono.fromCallable(() -> jdbcCall(str)).subscribeOn(fixedScheduler))
.flatMap(str -> asyncHttpCall(str))
.flatMap(str -> Mono.fromCallable(() -> jdbcCall(str)).subscribeOn(fixedScheduler))
.toFuture();
}
where fixedScheduler is used concurrently across HTTP requests.
We were hoping to get some feedback on this strategy for handling block calls within a decent number of fluxes. Of course, we understand that if all our requests were flowing through these blocking calls then we might as well not use reactor (outside of the admittedly nice processing API).
Update: Thanks bsideup for this answer. However, I should have been a little more specific with my questions.
My overall question is how to effectively have a blocking call used across multiple fluxes were these fluxes can be created/subscribed to in large numbers. We tried the suggested approach, but it results in an explosion of threads and quickly OOMs. So we are thinking to use a shared scheduler. So.. here are my questions.
Is using a shared scheduler (fixedScheduler) what you would suggest in the situation I describe? If not, will you point me in any directions?
If using a shared scheduler is good, would this be a good implementation of it: Schedulers.newParallel("blocking-scheduler", maxNumThreads)?
Update 2: Just dug a big on Schedulers#newParallel and realize that won't work since it 'rejects' blocking calls.
Really appreciate any tips!
While subscribeOn is indeed one way of handling blocking calls and your usage is okay, you can as well use publishOn.
It moves processing to the provided Scheduler, unless other publishOn is specified:
CompletionStage<String> getSomething(String input) {
return Mono.just(input)
.map(in -> process(in)) // process must be non-blocking, or go after publishOn
.publishOn(Schedulers.boundedElastic())
.map(::jdbcCall)
.flatMap(str -> asyncHttpCall(str))
.publishOn(Schedulers.boundedElastic())
.map(::jdbcCall)
.toFuture();
}
As you can see, you can continue using async calls too. Just make sure you're not blocking non-blocking schedulers (in that example, I use publishOn again after flatMap because asyncHttpCall may complete from non-blocking scheduler)
I have a gen_server process that registers a global name like this:
global:register_name(<<"CLIENT_", NAME/binary>>, self()),
Another process is trying to send this process a message using gen_server:call like this:
gen_server:call({global, <<"CLIENT_", NAME/binary>>}, {msg, DATA}),
If the second call happens before the first process registers the global name, it dies with:
exit with reason {noproc,{gen_server,call,[{global,<<"CLIENT_122">>},{msg, <<"TEST">>}]}}
What is the correct way to make a call only if the global name is register, and do something else if it is not?
Three things:
How to guard this call (mechanics).
Why you generally shouldn't want to guard the call (robust architecture).
Where you are putting your interface to this function (code structure).
Mechanics
You can check whether a name is registered with the global registry before making the call like this:
-spec send_message(Name, Message) -> Result
when Name :: term(),
Message :: term(),
Result :: {ok, term()}
| {error, no_proc}.
send_message(Name, Message) ->
case global:whereis_name(Name) of
undefined ->
{error, no_proc};
PID ->
Value = gen_server:call(PID, Message),
{ok, Value}
end.
Because there will be a few nanoseconds between the return value of global:whereis_name/1 being checked and the actual call via gen_server:call/2,3, however, so you still don't know if you actually just sent a call to a dead process, but at least you sent it to a PID that won't crash the program right away.
Another way to do it would be with a try ... catch construct, but that is a very tricky habit to get into.
Robust Architecture
All that stuff above, keep it in the back of your mind, but in the front of your mind you should want to crash if this name is unregistered. Your registered process is supposed to be alive so why are you being so paranoid?!? If thing are bad you want to know they are bad in a catastrophic way and let everything related to that crash and burn straight away. Don't try to recover on your own in an unknown state, that is what supervisors are for. Let your system be restarted in a known state and give it another go. If this is a user-directed action (some user of the system, or a web page request or whatever) they will try again because they are monkeys that try things more than once. If it is an automated request (the user is a computer or robot, for example) it can retry again or not, but leave that decision up to it in the common case -- but give it some indication of failure (an error message, a closed socket, etc.).
As long as the process you are calling registers its name during its init/1 call (before it has returned its own PID to its supervisor), and this is always happening before the calling process is alive or aware of the process to be called then you shouldn't have any trouble with this. If it has crashed for some reason, then you have more fundamental problems with your program and catching the caller's crash isn't going to help you. This is a basic idea in robustness engineering.
Structure your system so that the callee is guarantee to be alive and registered before the call can occur, and if it has died you should want the caller to die also. (Yes, I'm beating a dead horse, but this is important.)
Code Structure
Most of the time you don't want to have a module that defines a process, let's say foo.erl that defines a process we will name {global, "foo"}, have a naked call to gen_server:call/2,3 or gen_server:cast/2 that is intended for a separate process defined in another module (let's say bar.erl that defines a process we will name {global, "bar"}). What we would want is that bar.erl has an interface function that it exports, and that this function is where the gen_server:call/2 but happens.
That way any special work that applies to this call (which any other calling module may also require) exists in a single spot, and you can name the interface to the process "bar" in a way that conveys some meaning aside from the message being passed to it.
For example, if the process defined by bar.erl is a connection counter (maybe we are writing a game server and we're counting connections) we might have bar.erl be in charge of maintaining the counter. So processes send a cast (asynch message) to bar any time a new user connects. Instead of having every different process that might need to do this define some complex name checking and then naked message sending, instead consider having a function exported from bar.erl that hides that mess and is named something meaningful, like bar:notify_connect(). Just calling this in your other code is much easier to understand, and you can choose how you should be dealing with this "what if bar doesn't exist?" situation right there, in one spot.
On that note, you might want to take a look at the basic Erlang "service manager -> worker" pattern. Named processes are not overwhelmingly needed in many cases.
I am having a hard time wrapping my head around the correct way to make calls against a gen_server instance dynamically created by a supervisor with a simple_one_for_one child strategy. I am attempting to create data access controls as gen_servers. Each entity will have its own supervisor, and that supervisor will create gen_server instances as needed to actually perform CRUD operations on the database. I understand the process for defining the child processes, as well as the process for creating them as needed.
Initially, my plan was to abstract the child creation process into custom functions in the gen_server module that created a child, fired off the requested operation (e.g. find, store, delete) on that child using gen_server:call(), and then returning the operation results back to the calling process. Unless I am mistaken, though, that will block any other processes attempting to use those functions until the call returns. That is definitely not what I have in mind.
I may be stuck in OO mode (my background is Java), but it seems like there should be a clean way of allowing a function in one module to obtain a reference to a child process and then make calls against that process without leaking the internals of that child. In other words, I do not want to have to call the create_child() method on an entity supervisor and then have my application code make gen_server:calls against that child PID (i.e. gen_sever:call(Pid, {find_by_id, Id})). I would instead like to be able to call a function more like Child:find_by_id(Id).
A full answer is highly dependent on your application — for example, one gen_server might suffice, or you might really need a pool of database connections instead. But one thing you should be aware of is that a gen_server can return from a handle_call callback before it actually has a reply ready for the client by returning {noreply, NewState} and then later, once it has a client reply ready, calling gen_server:reply/2 to send it back to the client. This allows the gen_server to service calls from other clients without blocking on the first call. Note though that this requires that the gen_server has a way of sending a request into the database without having to block waiting for a reply; this is often achieved by having the database send a reply that arrives in the gen_server:handle_info/2 callback, passing enough info back that the gen_server can associate the database reply with the correct client request. Note also that gen_server:call/2,3 has a default timeout of 5 seconds, so you'll need to deal with that if you expect the duration of database calls to exceed the default.
when you create, modify or delete a record, you don't need to wait for an answer. You can use a gen_server:cast for this, but you don't need a gen_server for this, as I said in my first comment, a simple call to an interface function executed in the client process will save time.
If you want to read, 2 cases:
you can do something else while waiting the answer, then a gen_server call is ok, but a simple spawned process waiting for the answer and sending it back to the client will provide the same service.
you cannot do anything before getting the answer, then there is no blocking issue, and I think that it is really preferable to use as less code as possible so again a simple function call will be enough.
gen_server is meant to be persistent and react to messages. I don't see in your example the need to be persistent.
-module(access).
-export([add/2,get/1]).
-record(foo, {bar, baz}).
add(A,B) ->
F = fun() ->
mnesia:write(#foo{bar=A,baz=B})
end,
spawn(mnesia,activity,[transaction, F]). %% the function return immediately,
%% but you will not know if the transaction failed
get(Bar) ->
F = fun() ->
case mnesia:read({foo, Bar}) of
[#foo{baz=Baz}] -> Baz;
[] -> undefined
end
end,
Pid = self(),
Ref = make_ref(),
Get = fun() ->
R = mnesia:activity(transaction, F),
Pid ! {Ref,baz,R}
end,
spawn(Get),
Ref. %% the function return immediately a ref, and will send later the message {Ref,baz,Baz}.
If the problem you see is that you are leaking that the internal implementation of your db-process is a gen_server, you could implement the api such that it takes the pid as argument as well.
-module(user).
-behaviour(gen_server).
-export([find_by_id/2]).
find_by_id(Pid, Id) ->
gen_server:call(Pid, {find_by_id, Id}).
%% Lots of code omitted
handle_call({find_by_id, Id}, From, State) ->
ok.
%% Lots more code omitted.
This way you don't tell clients that the implementation is in fact a gen_server (although someone could use gen_server:call as well).
Suppose I have a gen_server that is handling only asynch calls (thus only handle_cast is implemented), should i keep handle_call and make it return only the generic ok value, or should i remove that part of the code and accept the warnings?
I'd opt for always returning {reply, ok, State}. Treat warnings as errors (that is, stop compilation and fix them). That way, when real warnings appear, they aren't hidden behind the noise.
It's a good habit to wrap all calls to behaviors in your own function, e.g:
delete(Something) ->
gen_server:call(?MODULE, {delete, Something}).
In this case, don't export any function which uses handle_call/2.
Isn't handling all of the calls a requirement for gen_server, regardless of whether you use them or not?
I noticed that messages sent to the pid of a gen_fsm process are matched in the state callbacks as events. Is this just accidental or can I rely on this feature?
Normally I would expect general messages sent to a gen_fsm to show up in the handle_info/3 callback and thought I would have to re-send it using gen_fsm:send_event.
Does gen_fsm try to match the message first to the state callback and then allways with the handle_info/3 callback? Or only if it doesn't match a state callback clause?
However when I try it my message seems to be handled twice according to debug output.
So basically the question can also be stated like: how to correctly handle received messages as events in gen_fsm state functions?
Clarification: that some of the events are occurring by getting messages passed should be considered given for this question.
I'm aware that in many cases its cleaner to make the protocol visible by using function calls into the fsm only.
I'm not so sure if this would improve the current framework where the mentioned gen_fsm has to fit in: Diverse protocol stacks where each layer calls a connect() function to attach (and sometimes start) the lower layer. Packets are sent to lower layers ba calling a function (send) and received by receiveing a message. Much like gen_tcp.
By looking at the code for gen_fsm I already figured out that general messages are only passed to handle_info, so only the question remains wether to call the state function directly from the handle_info/3 callback or resent using gen_fsm:send_event.
General messages are handled by handle_info callback, unless you have something like this in your code:
handle_info(Info, StateName, StateData) ->
?MODULE:StateName(Info, StateData).
Which avoids resending, but I do not recommend neither that, nor resending.
Delivering events exclusively by means of API calls encapsulating send_event/sync_send_event/send_all_state_event/sync_send_all_state_event makes protocol explicit. Which is a right thing, as it is easier to understand, maintain and document with edoc.