I'm building a library that others--i.e. those uninterested in internals--could use to pull data from our DBs. In the internals, I want a couple I/O calls to be performed in parallel for performance purposes. The trade-off here is that the client (who, again, might not care much about this whole threading thing) would need to provide an appropriate execution context. Therefore, I provide a suggestion to use a helpful execution context in a helper object:
object ThreadPoolHelper {
val cachedThreadPoolContext: ExecutionContext =
ExecutionContext.fromExecutor(Executors.newCachedThreadPool())
}
The question is (assuming that someday I also provide other options, like, say, a fixed thread pool for the clients to optionally use) am I fine just leaving this (these) as a val? Or am I better off making it lazy? Or a def?
One way or another, lazy is the way to go.
Making them lazy vals would be a good all-purpose choice, as each could be initialized as needed (as they are accessed). Then, you would never initialize more thread pools than are needed. scala.concurrent.ExecutionContext.Implicits.global is an implicit lazy val.
Technically, singleton objects (like ThreadPoolHelper) are lazy by default, so they will not be initialized until they are first accessed. A val would be fine if you only had one ExecutionContext in an object. However, multiple ExecutionContexts as vals in the same object wouldn't make as much sense, because accessing one would initialize them all--which would use more resources than needed.
A def would not make sense, because then you would be creating a new ExecutionContext on each call, and throwing it away when done. That could cause a lot of unwanted overhead, and default the purpose of having a thread pool in the first place.
Some ExecutionContexts that are little more custom than yours are singleton objects that extend ExecutionContext and implement their own custom behavior. These would also be lazy.
Related
Let's have a worker thread which is accessed from a wide variety of objects. This worker object has some public slots, so anyone who connects its signals to the worker's slots can use emit to trigger the worker thread's useful tasks.
This worker thread needs to be almost global, in the sense that several different classes use it, some of them are deep in the hierarchy (child of a child of a child of the main application).
I guess there are two major ways of doing this:
All the methods of the child classes pass their messages upwards the hierarchy via their return values, and let the main (e.g. the GUI) object handle all the emitting.
All those classes which require the services of the worker thread have a pointer to the Worker object (which is a member of the main class), and they all connect() to it in their constructors. Every such class then does the emitting by itself. Basically, dependency injection.
Option 2. seems much more clean and flexible to me, I'm only worried that it will create a huge number of connections. For example, if I have an array of an object which needs the thread, I will have a separate connection for each element of the array.
Is there an "official" way of doing this, as the creators of Qt intended it?
There is no magic silver bullet for this. You'll need to consider many factors, such as:
Why do those objects emit the data in the first place? Is it because they need to do something, that is, emission is a “command”? Then maybe they could call some sort of service to do the job without even worrying about whether it's going to happen in another thread or not. Or is it because they inform about an event? In such case they probably should just emit signals but not connect them. Its up to the using code to decide what to do with events.
How many objects are we talking about? Some performance tests are needed. Maybe it's not even an issue.
If there is an array of objects, what purpose does it serve? Perhaps instead of using a plain array some sort of “container” class is needed? Then the container could handle the emission and connection and objects could just do something like container()->handle(data). Then you'd only have one connection per container.
I'm having a go with greenDAO and so far it's going pretty well. One thing that doesn't seem to be covered by the docs or website (or anywhere :( ) is how it handles thread safety.
I know the basics mentioned elsewhere, like "use a single dao session" (general practice for Android + SQLite), and I understand the Java memory model quite well. The library internals even appear threadsafe, or at least built with that intention. But nothing I've seen covers this:
greenDAO caches entities by default. This is excellent for a completely single-threaded program - transparent and a massive performance boost for most uses. But if I e.g. loadAll() and then modify one of the elements, I'm modifying the same object globally across my app. If I'm using it on the main thread (e.g. for display), and updating the DB on a background thread (as is right and proper), there are obvious threading problems unless extra care is taken.
Does greenDAO do anything "under the hood" to protect against common application-level threading problems? For example, modifying a cached entity in the UI thread while saving it in a background thread (better hope they don't interleave! especially when modifying a list!)? Are there any "best practices" to protect against them, beyond general thread safety concerns (i.e. something that greenDAO expects and works well with)? Or is the whole cache fatally flawed from a multithreaded-application safety standpoint?
I've no experience with greenDAO but the documentation here:
http://greendao-orm.com/documentation/queries/
Says:
If you use queries in multiple threads, you must call forCurrentThread() on the query to get a Query instance for the current thread. Starting with greenDAO 1.3, object instances of Query are bound to their owning thread that build the query. This lets you safely set parameters on the Query object while other threads cannot interfere. If other threads try to set parameters on the query or execute the query bound to another thread, an exception will be thrown. Like this, you don’t need a synchronized statement. In fact you should avoid locking because this may lead to deadlocks if concurrent transactions use the same Query object.
To avoid those potential deadlocks completely, greenDAO 1.3 introduced the method forCurrentThread(). This will return a thread-local instance of the Query, which is safe to use in the current thread. Every time, forCurrentThread() is called, the parameters are set to the initial parameters at the time the query was built using its builder.
While so far as I can see the documentation doesn't explicitly say anything about multi threading other than this this seems pretty clear that it is handled. This is talking about multiple threads using the same Query object, so clearly multiple threads can access the same database. Certainly it's normal for databases and DAO to handle concurrent access and there are a lot of proven techniques for working with caches in this situation.
By default GreenDAO caches and returns cached entity instances to improve performance. To prevent this behaviour, you need to call:
daoSession.clear()
to clear all cached instances. Alternatively you can call:
objectDao.detachAll()
to clear cached instances only for the specific DAO object.
You will need to call these methods every time you want to clear the cached instances, so if you want to disable all caching, I recommend calling them in your Session or DAO accessor methods.
Documentation:
http://greenrobot.org/greendao/documentation/sessions/#Clear_the_identity_scope
Discussion: https://github.com/greenrobot/greenDAO/issues/776
Wanting to be sure we're using the correct synchronization (and no more than necessary) when writing threadsafe code in JRuby; specifically, in a Puma instantiated Rails app.
UPDATE: Extensively re-edited this question, to be very clear and use latest code we are implementing. This code uses the atomic gem written by #headius (Charles Nutter) for JRuby, but not sure it is totally necessary, or in which ways it's necessary, for what we're trying to do here.
Here's what we've got, is this overkill (meaning, are we over/uber-engineering this), or perhaps incorrect?
ourgem.rb:
require 'atomic' # gem from #headius
SUPPORTED_SERVICES = %w(serviceABC anotherSvc andSoOnSvc).freeze
module Foo
def self.included(cls)
cls.extend(ClassMethods)
cls.send :__setup
end
module ClassMethods
def get(service_name, method_name, *args)
__cached_client(service_name).send(method_name.to_sym, *args)
# we also capture exceptions here, but leaving those out for brevity
end
private
def __client(service_name)
# obtain and return a client handle for the given service_name
# we definitely want to cache the value returned from this method
# **AND**
# it is a requirement that this method ONLY be called *once PER service_name*.
end
def __cached_client(service_name)
##_clients.value[service_name]
end
def __setup
##_clients = Atomic.new({})
##_clients.update do |current_service|
SUPPORTED_SERVICES.inject(Atomic.new({}).value) do |memo, service_name|
if current_services[service_name]
current_services[service_name]
else
memo.merge({service_name => __client(service_name)})
end
end
end
end
end
end
client.rb:
require 'ourgem'
class GetStuffFromServiceABC
include Foo
def self.get_some_stuff
result = get('serviceABC', 'method_bar', 'arg1', 'arg2', 'arg3')
puts result
end
end
Summary of the above: we have ##_clients (a mutable class variable holding a Hash of clients) which we only want to populate ONCE for all available services, which are keyed on service_name.
Since the hash is in a class variable (and hence threadsafe?), are we guaranteed that the call to __client will not get run more than once per service name (even if Puma is instantiating multiple threads with this class to service all the requests from different users)? If the class variable is threadsafe (in that way), then perhaps the Atomic.new({}) is unnecessary?
Also, should we be using an Atomic.new(ThreadSafe::Hash) instead? Or again, is that not necessary?
If not (meaning: you think we do need the Atomic.news at least, and perhaps also the ThreadSafe::Hash), then why couldn't a second (or third, etc.) thread interrupt between the Atomic.new(nil) and the ##_clients.update do ... meaning the Atomic.news from EACH thread will EACH create two (separate) objects?
Thanks for any thread-safety advice, we don't see any questions on SO that directly address this issue.
Just a friendly piece of advice, before I attempt to tackle the issues you raise here:
This question, and the accompanying code, strongly suggests that you don't (yet) have a solid grasp of the issues involved in writing multi-threaded code. I encourage you to think twice before deciding to write a multi-threaded app for production use. Why do you actually want to use Puma? Is it for performance? Will your app handle many long-running, I/O-bound requests (like uploading/downloading large files) at the same time? Or (like many apps) will it primarily handle short, CPU-bound requests?
If the answer is "short/CPU-bound", then you have little to gain from using Puma. Multiple single-threaded server processes would be better. Memory consumption will be higher, but you will keep your sanity. Writing correct multi-threaded code is devilishly hard, and even experts make mistakes. If your business success, job security, etc. depends on that multi-threaded code working and working right, you are going to cause yourself a lot of unnecessary pain and mental anguish.
That aside, let me try to unravel some of the issues raised in your question. There is so much to say that it's hard to know where to start. You may want to pour yourself a cold or hot beverage of your choice before sitting down to read this treatise:
When you talk about writing "thread-safe" code, you need to be clear about what you mean. In most cases, "thread-safe" code means code which doesn't concurrently modify mutable data in a way which could cause data corruption. (What a mouthful!) That could mean that the code doesn't allow concurrent modification of mutable data at all (using locks), or that it does allow concurrent modification, but makes sure that it doesn't corrupt data (probably using atomic operations and a touch of black magic).
Note that when your threads are only reading data, not modifying it, or when working with shared stateless objects, there is no question of "thread safety".
Another definition of "thread-safe", which probably applies better to your situation, has to do with operations which affect the outside world (basically I/O). You may want some operations to only happen once, or to happen in a specific order. If the code which performs those operations runs on multiple threads, they could happen more times than desired, or in a different order than desired, unless you do something to prevent that.
It appears that your __setup method is only called when ourgem.rb is first loaded. As far as I know, even if multiple threads require the same file at the same time, MRI will only ever let a single thread load the file. I don't know whether JRuby is the same. But in any case, if your source files are being loaded more than once, that is symptomatic of a deeper problem. They should only be loaded once, on a single thread. If your app handles requests on multiple threads, those threads should be started up after the application has loaded, not before. This is the only sane way to do things.
Assuming that everything is sane, ourgem.rb will be loaded using a single thread. That means __setup will only ever be called by a single thread. In that case, there is no question of thread safety at all to worry about (as far as initialization of your "client cache" goes).
Even if __setup was to be called concurrently by multiple threads, your atomic code won't do what you think it does. First of all, you use Atomic.new({}).value. This wraps a Hash in an atomic reference, then unwraps it so you just get back the Hash. It's a no-op. You could just write {} instead.
Second, your Atomic#update call will not prevent the initialization code from running more than once. To understand this, you need to know what Atomic actually does.
Let me pull out the old, tired "increment a shared counter" example. Imagine the following code is running on 2 threads:
i += 1
We all know what can go wrong here. You may end up with the following sequence of events:
Thread A reads i and increments it.
Thread B reads i and increments it.
Thread A writes its incremented value back to i.
Thread B writes its incremented value back to i.
So we lose an update, right? But what if we store the counter value in an atomic reference, and use Atomic#update? Then it would be like this:
Thread A reads i and increments it.
Thread B reads i and increments it.
Thread A tries to write its incremented value back to i, and succeeds.
Thread B tries to write its incremented value back to i, and fails, because the value has already changed.
Thread B reads i again and increments it.
Thread B tries to write its incremented value back to i again, and succeeds this time.
Do you get the idea? Atomic never stops 2 threads from running the same code at the same time. What it does do, is force some threads to retry the #update block when necessary, to avoid lost updates.
If your goal is to ensure that your initialization code will only ever run once, using Atomic is a very inappropriate choice. If anything, it could make it run more times, rather than less (due to retries).
So, that is that. But if you're still with me here, I am actually more concerned about whether your "client" objects are themselves thread-safe. Do they have any mutable state? Since you are caching them, it seems that initializing them must be slow. Be that as it may, if you use locks to make them thread-safe, you may not be gaining anything from caching and sharing them between threads. Your "multi-threaded" server may be reduced to what is effectively an unnecessarily complicated, single-threaded server.
If the client objects have no mutable state, good for you. You can be "free and easy" and share them between threads with no problems. If they do have mutable state, but initializing them is slow, then I would recommend caching one object per thread, so they are never shared. Thread[] is your friend there.
Actually I am using visual C++ to try to bind lua functions as callbacks for socket events(in another thread). I initialize the lua stuff in one thread and the socket is in another thread, so every time the socket sends/receives a message, it will call the lua function and the lua function determines what it should do according to the 'tag' within the message.
So my questions are:
Since I pass the same Lua state to lua functions, is that safe? Doesn't it need some kinda protection? The lua functions are called from another thead so I guess they might be called simultaneously.
If it is not safe, what's the solution for this case?
It is not safe to call back asynchronously into a Lua state.
There are many approaches to dealing with this. The most popular involve some kind of polling.
A recent generic synchronization library is DarkSideSync
A popular Lua binding to libev is lua-ev
This SO answer recommends Lua Lanes with LuaSocket.
It is not safe to call function within one Lua state simultaneously in multiple threads.
I was dealing with the same problem, since in my application all basics such as communication are handled by C++ and all the business logic is implemented in Lua. What I do is create a pool of Lua states that are all created and initialised on an incremental basis (once there's not enough states, create one and initialise with common functions / objects). It works like this:
Once a connection thread needs to call a Lua function, it checks out an instance of Lua state, initialises specific globals (I call it a thread / connection context) in a separate (proxy) global table that prevents polluting the original global, but is indexed by the original global
Call a Lua function
Check the Lua state back in to the pool, where it is restored to the "ready" state (dispose of the proxy global table)
I think this approach would be well suited for your case as well. The pool checks each state (on an interval basis) when it was last checked out. When the time difference is big enough, it destroys the state to preserve resources and adjust the number of active states to current server load. The state that is checked out is the most recently used among the available states.
There are some things you need to consider when implementing such a pool:
Each state needs to be populated with the same variables and global functions, which increases memory consumption.
Implementing an upper limit for state count in the pool
Ensuring all the globals in each state are in a consistent state, if they happen to change (here I would recommend prepopulating only static globals, while populating dynamic ones when checking out a state)
Dynamic loading of functions. In my case there are many thousands of functions / procedures that can be called in Lua. Having them constantly loaded in all states would be a huge waste. So instead I keep them byte code compiled on the C++ side and have them loaded when needed. It turns out not to impact performance that much in my case, but your mileage may vary. One thing to keep in mind is to load them only once. Say you invoke a script that needs to call another dynamically loaded function in a loop. Then you should load the function as a local once before the loop. Doing it otherwise would be a huge performance hit.
Of course this is just one idea, but one that turned out to be best suited for me.
It's not safe, as the others mentioned
Depends on your usecase
Simplest solution is using a global lock using the lua_lock and lua_unlock macros. That would use a single Lua state, locked by a single mutex. For a low number of callbacks it might suffice, but for higher traffic it probably won't due to the overhead incurred.
Once you need better performance, the Lua state pool as mentioned by W.B. is a nice way to handle this. Trickiest part here I find synchronizing the global data across the multiple states.
DarkSideSync, mentioned by Doug, is useful in cases where the main application loop resides on the Lua side. I specifically wrote it for that purpose. In your case this doesn't seem a fit. Having said that; depending on your needs, you might consider changing your application so the main loop does reside on the Lua side. If you only handle sockets, then you can use LuaSocket and no synchronization is required at all. But obviously that depends on what else the application does.
I'm wondering if it is possible to add a member object that can be used across multiple map() call. For example, a StringBuilder:
private StringBuilder builder;
public void map(...){
...
builder.setLength(0);
builder.append(a);
builder.append(b);
builder.append(c);
d = builder.toString();
...
}
Obviously, if the mapper object is shared across multiple threads, the builder object above will not behave as expected due to concurrent access from more than one threads.
So my question is: Is it assured that each thread in hadoop will use one dedicated mapper object for itself? Or it is a configurable behavior?
Thanks
As long as you are not using the MultithreadedMapper class, but your own, there will be no problem. map() is called sequential and not in parallel.
It is common to use a StringBuilder or other data structures to buffer a few objects between the calls.
But make sure you clone the objects from your input objects, there is only one object and it will be filled over and over again to prevent lots of GC.
So there is no need to synchronize or take care of race conditions.
I don't think that's possible. The reason for that is that each mapper runs in its own JVM (they will be distributed on different machines), so there's no way you can share a variable or object across multiple mappers or reducers easily.
Now if all your mappers run on the same node, I believe there is a configuration for JVM reuse somewhere, but honestly I wouldn't bother with that, especially if all you need is a StringBuilder :)
I've seen this question once before, and it could be solved very easily by changing the design of the application. Maybe you can tell more about what you're trying to accomplish with this to see if this is really needed. If you really need it, you can still serialize your object, put it in HDFS, then read it with each mapper, deserialize it, but that seems backwards.