Calling Aws::initAPI() spawns multiple threads

[dbowman-ion]

It appears that a call to Aws::initAPI() spawns '# cpu cores / 2' threads but so far I've been unable to find any documentation on how or why this happens, or how to configure it.

We spawn multiple process that call Aws::InitAPI and on a high cpu count machine this results in more than 2000 threads spawned by the SDK. As threads/PID's are not an infinite resource, any advice on how to remove/limit this behaviour would be appreciated.

Minimal reproducable example:

#include <thread>
#include <chrono>
#include <aws/core/Aws.h>
int main(int, char**)
{
    Aws::SDKOptions options;
    Aws::InitAPI(options);
    std::this_thread::sleep_for( std::chrono::seconds( 600 ) );
    return 0;
}

Compile & run and then do a pstack on its pid, you'll see multiple threads like:

Thread 19 (Thread 0x7f70257e2700 (LWP 301201)):
#0  0x00007f70349fc923 in epoll_wait () from /lib64/libc.so.6
#1  0x00007f7035b73c1b in s_main_loop () from /path/to/aws/1.9.114/lib64/libaws-cpp-sdk-core.so
#2  0x00007f7035c2ed5f in thread_fn () from /path/to/aws/1.9.114/lib64/libaws-cpp-sdk-core.so
#3  0x00007f7034ccee25 in start_thread () from /lib64/libpthread.so.0
#4  0x00007f70349fc34d in clone () from /lib64/libc.so.6

[dbowman-ion]

Did a bit more investigation, and was able to limit the threads via:

    options.ioOptions.clientBootstrap_create_fn = []()
    {
        Aws::Crt::Io::EventLoopGroup eventLoopGroup( 2 );
        Aws::Crt::Io::DefaultHostResolver defaultHostResolver(eventLoopGroup, 8, 30);
        auto clientBootstrap = Aws::MakeShared<Aws::Crt::Io::ClientBootstrap>(ALLOCATION_TAG, eventLoopGroup, defaultHostResolver);
        clientBootstrap->EnableBlockingShutdown();
        return clientBootstrap;
    };

[KaibaLopez]

Hi @dbowman-ion ,
So.. it seems you've found a sort of workaround, it this now a question on why this happens or?

[github-actions]

Greetings! It looks like this issue hasn’t been active in longer than a week. We encourage you to check if this is still an issue in the latest release. Because it has been longer than a week since the last update on this, and in the absence of more information, we will be closing this issue soon. If you find that this is still a problem, please feel free to provide a comment or add an upvote to prevent automatic closure, or if the issue is already closed, please feel free to open a new one.

[dbowman-ion]

Yes, understanding why would be nice.
It may also be worth some documentation as neither of these mention anything about threads or how to limit them:
https://sdk.amazonaws.com/cpp/api/LATEST/struct_aws_1_1_s_d_k_options.html
https://sdk.amazonaws.com/cpp/api/LATEST/struct_aws_1_1_io_options.html

I ended up finding the answer in aws-sdk-cpp-core/source/Aws.cpp

[saad-alsaad1]

Hi,

I guess I have the same issue in my case.
I have an executable that gets some objects from S3, when I call the executable multiple times (sequential about 20-30 times) I get the crash below.
The crash happens when calling Aws::initAPI()
code:

#include <aws/core/Aws.h>

static Aws::SDKOptions options;
void init_aws(){
Aws::InitAPI(options);
}

main(){
init_aws();
Aws::ShutdownAPI(options);
}

Trace:
=5629==ERROR: AddressSanitizer: SEGV on unknown address 0x7fff65800002 (pc 0x7f3868b462bd bp 0x611000000400 sp 0x7f3861bfb0c0 T1)
==5629==The signal is caused by a READ memory access.
#0 0x7f3868b462bc in getenv (/lib64/libc.so.6+0x392bc)
#1 0x7f3867b5f863 (/path/libuv.so.1+0xc863)
#2 0x7f386993820a in __pthread_once_slow (/lib64/libpthread.so.0+0x620a)
#3 0x7f3867b71058 in uv_once (/path/libuv.so.1+0x1e058)
#4 0x7f3867b5fab8 in uv__work_submit (/path/libuv.so.1+0xcab8)
#5 0x7f3867b6a5a1 in uv_getaddrinfo (/path/libuv.so.1+0x175a1)

[pgj007]

This problem occurs when the InitAPI and ShutdownAPI are not in the same thread。