Conor  Grady

Conor Grady

1664919060

Fast, Scalable, Self-contained, Single-threaded Java Web Server

Microhttp

Microhttp is a fast, scalable, event-driven, self-contained Java web server that is small enough for a programmer to understand and reason about. It does not rely on any classpath dependencies or native code.

It is capable of serving over 1,000,000 requests per second on a commodity EC2 host (c5.2xlarge). TechEmpower continuous benchmarking results consistently show Microhttp achieves over 2,000,000 requests per second.

Comprehensibility is the highest priority. This library is intended to be an alternative to commonly used frameworks with overwhelming complexity.

Microhttp discretizes all requests and responses. Streaming is not supported. This aligns well with transactional web services that exchange small payloads.

Microhttp supports aspects of HTTP 1.0 and HTTP 1.1, but it is not fully compliant with the spec (RFC 2616, RFC 7230, etc.) 100-Continue (RFC 2616 8.2.3) is not implemented, for example.

TLS is not supported. Edge proxies and load balancers provide this capability. The last hop to Microhttp typically does not require TLS.

HTTP 2 is not supported for a similar reason. Edge proxies can support HTTP 2 while using HTTP 1.1 on the last hop to Microhttp.

Microhttp is 100% compatible with Project Loom virtual threads. Simply handle each request in a separate virtual thread, invoking the callback function upon completion.

Principles:

  • No dependencies
  • Small, targeted codebase (~500 LOC)
  • Highly concurrent
  • Single-threaded event loops
  • Event-driven non-blocking NIO
  • No TLS support
  • No streaming support
  • Traceability via log events

Includes:

  • HTTP 1.0 and 1.1
  • Chunked transfer encoding
  • Persistent connections
  • Pipelining

Intended Use:

  • Teaching or learning scalable concurrency, NIO, HTTP, networking
  • Mock or stub servers for testing
  • Internal web servers not exposed to the internet
  • Web server behind an internet-facing reverse proxy (Nginx, HAProxy, AWS ELB, etc)

Dependency

Microhttp is available in the Maven Central repository with group org.microhttp and artifact microhttp.

<dependency>
    <groupId>org.microhttp</groupId>
    <artifactId>microhttp</artifactId>
    <version>0.8</version>
</dependency>

Getting Started

The snippet below represents a minimal starting point. Default options and debug logging.

The application consists of an event loop running in a background thread.

Responses are handled immediately in the Handler.handle method.

Response response = new Response(
        200,
        "OK",
        List.of(new Header("Content-Type", "text/plain")),
        "hello world\n".getBytes());
Handler handler = (req, callback) -> callback.accept(response);
EventLoop eventLoop = new EventLoop(handler);
eventLoop.start();
eventLoop.join();

The following example demonstrates the full range of configuration options.

Response response = new Response(
        200,
        "OK",
        List.of(new Header("Content-Type", "text/plain")),
        "hello world\n".getBytes());
Options options = new Options()
        .withHost("localhost")
        .withPort(8080)
        .withRequestTimeout(Duration.ofSeconds(60))
        .withResolution(Duration.ofMillis(100))
        .withBufferSize(1_024 * 64)
        .withMaxRequestSize(1_024 * 1_024)
        .withAcceptLength(0)
        .withConcurrency(4);
Logger logger = new DebugLogger();
Handler handler = (req, callback) -> callback.accept(response);
EventLoop eventLoop = new EventLoop(options, logger, handler);
eventLoop.start();
eventLoop.join();

The example below demonstrates asynchronous request handling.

Responses are handled in a separate background thread after an artificial one-second delay.

Response response = new Response(
        200,
        "OK",
        List.of(new Header("Content-Type", "text/plain")),
        "hello world\n".getBytes());
ScheduledExecutorService executorService = Executors.newSingleThreadScheduledExecutor();
Handler handler = (req, callback) -> executorService.schedule(() -> callback.accept(response), 1, TimeUnit.SECONDS);
EventLoop eventLoop = new EventLoop(handler);
eventLoop.start();
eventLoop.join();

Benchmarks

These benchmark were performed on July 12, 2022 with commit 78f54e84e86cdd038c87baaf45b7973a8f088cf7.

The experiments detailed below were conducted on a pair of EC2 instances in AWS, one running the server and another running the client.

  • Region: us-west-2
  • Instance type: c5.2xlarge compute optimized instance 8 vCPU and 16 GB of memory
  • OS: Amazon Linux 2 with Linux Kernel 5.10, AMI ami-00f7e5c52c0f43726
  • OpenJDK 18.0.1.1 from https://jdk.java.net/18/

The wrk HTTP benchmarking tool was used to generate load on the client EC2 instance.

Throughput

The goal of throughput benchmarks is to gauge the maximum request-per-second rate that can be supported by Microhttp. These experiments are intended to surface the costs and limitations of Microhttp alone. They are not intended to provide a real world estimate of throughput in an integrated system with many components and dependencies.

Server

ThroughputServer.java was used for throughput tests.

It simply returns "hello world" in a tiny, plain-text response to every request. Requests are handled in the context of the event loop thread, directly within the Handler.handle method.

./jdk-18.0.1.1/bin/java -cp microhttp-0.8-SNAPSHOT.jar org.microhttp.ThroughputServer

Benchmark

With HTTP pipelining, a request rate of over 1,000,000 requests per second was consistently reproducible.

In the 1-minute run below, a rate of 1,098,810 requests per second was achieved.

  • 100 concurrent connections
  • 1 wrk worker threads
  • 10 second timeout
  • 16 pipelined requests

No custom kernel parameters were set beyond the AMI defaults for this test.

No errors occurred and the 99th percentile response time was quite reasonable, given that client and server were both CPU-bound.

$ date
Tue Jul 12 17:11:05 UTC 2022

$ ./wrk -H "Host: 10.39.196.71:8080" -H "Accept: text/plain" -H "Connection: keep-alive" --latency -d 60s -c 100 --timeout 10 -t 1 http://10.39.196.71:8080/ -s pipeline.lua -- 16
Running 1m test @ http://10.39.196.71:8080/
  1 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    18.44ms   13.95ms  52.12ms   53.25%
    Req/Sec     1.10M    22.87k    1.14M    87.83%
  Latency Distribution
     50%   18.37ms
     75%   31.47ms
     90%   39.33ms
     99%    0.00us
  65929433 requests in 1.00m, 4.73GB read
Requests/sec: 1098810.79
Transfer/sec:     80.69MB

Without HTTP pipelining, a request rate of over 450,000 requests per second was consistently reproducible.

In the 1-minute run below, a rate of 454,796 requests per second was achieved.

  • 100 concurrent connections
  • 8 wrk worker threads
  • 10 second timeout

No errors occurred and the 99th percentile response time was exceptional.

$ date
Tue Jul 12 17:16:49 UTC 2022
 
$ ./wrk -H "Host: 10.39.196.71:8080" -H "Accept: text/plain" -H "Connection: keep-alive" --latency -d 60s -c 100 --timeout 10 -t 8 http://10.39.196.71:8080/
Running 1m test @ http://10.39.196.71:8080/
  8 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   218.65us    1.64ms 212.93ms   99.97%
    Req/Sec    57.15k     4.68k   69.47k    85.19%
  Latency Distribution
     50%  188.00us
     75%  229.00us
     90%  277.00us
     99%  372.00us
  27332950 requests in 1.00m, 1.96GB read
Requests/sec: 454796.26
Transfer/sec:     33.40MB

Concurrency

The goal of concurrency benchmarks is to gauge the number of concurrent connections and clients that can be supported by Microhttp.

In order to facilitate the rapid creation of 50,000 connections, the following sysctl kernel parameter changes were committed on both hosts prior to the start of the experiment:

sysctl net.ipv4.ip_local_port_range="2000 64000"
sysctl net.ipv4.tcp_fin_timeout=30
sysctl net.core.somaxconn=8192
sysctl net.core.netdev_max_backlog=8000
sysctl net.ipv4.tcp_max_syn_backlog=8192

Server

ConcurrencyServer.java was used for concurrency tests.

"hello world" responses are handled in a separate background thread after an injected one-second delay. The one-second delay dramatically reduces the resource footprint since requests and responses aren't speeding over each connection continuously. This leaves room to scale up connections, which is the metric of interest.

./jdk-18.0.1.1/bin/java -cp microhttp-0.8-SNAPSHOT.jar org.microhttp.ConcurrencyServer 8192

Benchmark

A concurrency level of 50,000 connections without error was consistently reproducible.

  • 50,000 concurrent connections
  • 16 wrk worker threads
  • 10 second timeout

No errors occurred.

The quality of service is stellar. The 99% percentile response time it 1.01 seconds, just 0.01 above the target latency introduced on the server.

$ date
Tue Jul 12 17:26:53 UTC 2022

$ ./wrk -H "Host: 10.39.196.71:8080" -H "Accept: text/plain" -H "Connection: keep-alive" --latency -d 60s -c 50000 --timeout 10 -t 16 http://10.39.196.71:8080/
Running 1m test @ http://10.39.196.71:8080/
  16 threads and 50000 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency     1.00s     2.74ms   1.21s    95.44%
    Req/Sec     8.52k    11.02k   31.56k    73.64%
  Latency Distribution
     50%    1.00s 
     75%    1.00s 
     90%    1.00s 
     99%    1.01s 
  2456381 requests in 1.00m, 180.38MB read
Requests/sec:  40875.87
Transfer/sec:      3.00MB

Download Details:

Author: ebarlas
Source Code: https://github.com/ebarlas/microhttp

License: MIT license

#java 

What is GEEK

Buddha Community

Fast, Scalable, Self-contained, Single-threaded Java Web Server

Pyringe: Debugger Capable Of Attaching to & Injecting Code Into Python

DISCLAIMER: This is not an official google project, this is just something I wrote while at Google.

Pyringe

What this is

Pyringe is a python debugger capable of attaching to running processes, inspecting their state and even of injecting python code into them while they're running. With pyringe, you can list threads, get tracebacks, inspect locals/globals/builtins of running functions, all without having to prepare your program for it.

What this is not

A "Google project". It's my internship project that got open-sourced. Sorry for the confusion.

What do I need?

Pyringe internally uses gdb to do a lot of its heavy lifting, so you will need a fairly recent build of gdb (version 7.4 onwards, and only if gdb was configured with --with-python). You will also need the symbols for whatever build of python you're running.
On Fedora, the package you're looking for is python-debuginfo, on Debian it's called python2.7-dbg (adjust according to version). Arch Linux users: see issue #5, Ubuntu users can only debug the python-dbg binary (see issue #19).
Having Colorama will get you output in boldface, but it's optional.

How do I get it?

Get it from the Github repo, PyPI, or via pip (pip install pyringe).

Is this Python3-friendly?

Short answer: No, sorry. Long answer:
There's three potentially different versions of python in play here:

  1. The version running pyringe
  2. The version being debugged
  3. The version of libpythonXX.so your build of gdb was linked against

2 Is currently the dealbreaker here. Cpython has changed a bit in the meantime[1], and making all features work while debugging python3 will have to take a back seat for now until the more glaring issues have been taken care of.
As for 1 and 3, the 2to3 tool may be able to handle it automatically. But then, as long as 2 hasn't been taken care of, this isn't really a use case in the first place.

[1] - For example, pendingbusy (which is used for injection) has been renamed to busy and been given a function-local scope, making it harder to interact with via gdb.

Will this work with PyPy?

Unfortunately, no. Since this makes use of some CPython internals and implementation details, only CPython is supported. If you don't know what PyPy or CPython are, you'll probably be fine.

Why not PDB?

PDB is great. Use it where applicable! But sometimes it isn't.
Like when python itself crashes, gets stuck in some C extension, or you want to inspect data without stopping a program. In such cases, PDB (and all other debuggers that run within the interpreter itself) are next to useless, and without pyringe you'd be left with having to debug using print statements. Pyringe is just quite convenient in these cases.

I injected a change to a local var into a function and it's not showing up!

This is a known limitation. Things like inject('var = 2') won't work, but inject('var[1] = 1337') should. This is because most of the time, python internally uses a fast path for looking up local variables that doesn't actually perform the dictionary lookup in locals(). In general, code you inject into processes with pyringe is very different from a normal python function call.

How do I use it?

You can start the debugger by executing python -m pyringe. Alternatively:

import pyringe
pyringe.interact()

If that reminds you of the code module, good; this is intentional.
After starting the debugger, you'll be greeted by what behaves almost like a regular python REPL.
Try the following:

==> pid:[None] #threads:[0] current thread:[None]
>>> help()
Available commands:
 attach: Attach to the process with the given pid.
 bt: Get a backtrace of the current position.
 [...]
==> pid:[None] #threads:[0] current thread:[None]
>>> attach(12679)
==> pid:[12679] #threads:[11] current thread:[140108099462912]
>>> threads()
[140108099462912, 140108107855616, 140108116248323, 140108124641024, 140108133033728, 140108224739072, 140108233131776, 140108141426432, 140108241524480, 140108249917184, 140108269324032]

The IDs you see here correspond to what threading.current_thread().ident would tell you.
All debugger functions are just regular python functions that have been exposed to the REPL, so you can do things like the following.

==> pid:[12679] #threads:[11] current thread:[140108099462912]
>>> for tid in threads():
...   if not tid % 10:
...     thread(tid)
...     bt()
... 
Traceback (most recent call last):
  File "/usr/lib/python2.7/threading.py", line 524, in __bootstrap
    self.__bootstrap_inner()
  File "/usr/lib/python2.7/threading.py", line 551, in __bootstrap_inner
    self.run()
  File "/usr/lib/python2.7/threading.py", line 504, in run
    self.__target(*self.__args, **self.__kwargs)
  File "./test.py", line 46, in Idle
    Thread_2_Func(1)
  File "./test.py", line 40, in Wait
    time.sleep(n)
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> 

You can access the inferior's locals and inspect them like so:

==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> inflocals()
{'a': <proxy of A object at remote 0x1d9b290>, 'LOL': 'success!', 'b': <proxy of B object at remote 0x1d988c0>, 'n': 1}
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> p('a')
<proxy of A object at remote 0x1d9b290>
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> p('a').attr
'Some_magic_string'
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> 

And sure enough, the definition of a's class reads:

class Example(object):
  cl_attr = False
  def __init__(self):
    self.attr = 'Some_magic_string'

There's limits to how far this proxying of objects goes, and everything that isn't trivial data will show up as strings (like '<function at remote 0x1d957d0>').
You can inject python code into running programs. Of course, there are caveats but... see for yourself:

==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> inject('import threading')
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> inject('print threading.current_thread().ident')
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> 

The output of my program in this case reads:

140108241524480

If you need additional pointers, just try using python's help (pyhelp() in the debugger) on debugger commands.

Author: google
Source Code: https://github.com/google/pyringe
License: Apache-2.0 License

#python 

Libraries for Debugging Code in Popular Python

In this Python article, let's learn about Debugging Tools: Libraries for Debugging Code in Popular Python

Table of contents:

  • pdb-like Debugger
    • ipdb - IPython-enabled pdb.
    • pdb++ - Another drop-in replacement for pdb.
    • pudb - A full-screen, console-based Python debugger.
    • wdb - An improbable web debugger through WebSockets.
  • Tracing
    • lptrace - strace for Python programs.
    • manhole - Debugging UNIX socket connections and present the stacktraces for all threads and an interactive prompt.
    • pyringe - Debugger capable of attaching to and injecting code into Python processes.
    • python-hunter - A flexible code tracing toolkit.
  • Profiler
    • line_profiler - Line-by-line profiling.
    • memory_profiler - Monitor Memory usage of Python code.
    • py-spy - A sampling profiler for Python programs. Written in Rust.
    • pyflame - A ptracing profiler For Python.
    • vprof - Visual Python profiler.
  • Others
    • django-debug-toolbar - Display various debug information for Django.
    • django-devserver - A drop-in replacement for Django's runserver.
    • flask-debugtoolbar - A port of the django-debug-toolbar to flask.
    • icecream - Inspect variables, expressions, and program execution with a single, simple function call.
    • pyelftools - Parsing and analyzing ELF files and DWARF debugging information.

 

What is a debugging tool?

A debugger is a software tool that can help the software development process by identifying coding errors at various stages of the operating system or application development. Some debuggers will analyze a test run to see what lines of code were not executed.

Debugger for Python programs with a graphical user interface. It uses bdb (part of stdlib) but adds a GUI and has some powerful features like object browser, windows for variables, classes, functions, exceptions, stack, conditional breakpoints, etc.


Libraries for Debugging Code in Popular Python

  1. IPython pdb

ipdb exports functions to access the IPython debugger, which features tab completion, syntax highlighting, better tracebacks, better introspection with the same interface as the pdb module.

Example usage:

import ipdb
ipdb.set_trace()
ipdb.set_trace(context=5)  # will show five lines of code
                           # instead of the default three lines
                           # or you can set it via IPDB_CONTEXT_SIZE env variable
                           # or setup.cfg file
ipdb.pm()
ipdb.run('x[0] = 3')
result = ipdb.runcall(function, arg0, arg1, kwarg='foo')
result = ipdb.runeval('f(1,2) - 3')

Arguments for set_trace

The set_trace function accepts context which will show as many lines of code as defined, and cond, which accepts boolean values (such as abc == 17) and will start ipdb's interface whenever cond equals to True.

Using configuration file

It's possible to set up context using a .ipdb file on your home folder, setup.cfg or pyproject.toml on your project folder. You can also set your file location via env var $IPDB_CONFIG. Your environment variable has priority over the home configuration file, which in turn has priority over the setup config file. Currently, only context setting is available.

A valid setup.cfg is as follows

[ipdb]
context=5

A valid .ipdb is as follows

context=5

A valid pyproject.toml is as follows

[tool.ipdb]
context=5

The post-mortem function, ipdb.pm(), is equivalent to the magic function %debug.

View on GitHub


2.  pdb++

pdb++, a drop-in replacement for pdb (the Python debugger)

What is it?

This module is an extension of the pdb module of the standard library. It is meant to be fully compatible with its predecessor, yet it introduces a number of new features to make your debugging experience as nice as possible.

https://user-images.githubusercontent.com/412005/64484794-2f373380-d20f-11e9-9f04-e1dabf113c6f.png

pdb++ features include:

  • colorful TAB completion of Python expressions (through fancycompleter)
  • optional syntax highlighting of code listings (through Pygments)
  • sticky mode
  • several new commands to be used from the interactive (Pdb++) prompt
  • smart command parsing (hint: have you ever typed r or c at the prompt to print the value of some variable?)
  • additional convenience functions in the pdb module, to be used from your program

pdb++ is meant to be a drop-in replacement for pdb. If you find some unexpected behavior, please report it as a bug.

Installation

Since pdb++ is not a valid package name the package is named pdbpp:

$ pip install pdbpp

pdb++ is also available via conda:

$ conda install -c conda-forge pdbpp

Alternatively, you can just put pdb.py somewhere inside your PYTHONPATH.

View on GitHub


3.  PuDB

Its goal is to provide all the niceties of modern GUI-based debuggers in a more lightweight and keyboard-friendly package. PuDB allows you to debug code right where you write and test it--in a terminal.

Here are some screenshots:

Light theme

  • doc/images/pudb-screenshot-light.png

Dark theme

  • doc/images/pudb-screenshot-dark.png

View on GitHub


4.  wdb

An improbable web debugger through WebSockets

wdb is a full featured web debugger based on a client-server architecture.

The wdb server which is responsible of managing debugging instances along with browser connections (through websockets) is based on Tornado. The wdb clients allow step by step debugging, in-program python code execution, code edition (based on CodeMirror) setting breakpoints...

Due to this architecture, all of this is fully compatible with multithread and multiprocess programs.

wdb works with python 2 (2.6, 2.7), python 3 (3.2, 3.3, 3.4, 3.5) and pypy. Even better, it is possible to debug a python 2 program with a wdb server running on python 3 and vice-versa or debug a program running on a computer with a debugging server running on another computer inside a web page on a third computer!

Even betterer, it is now possible to pause a currently running python process/thread using code injection from the web interface. (This requires gdb and ptrace enabled)

In other words it's a very enhanced version of pdb directly in your browser with nice features.

Installation:

Global installation:

    $ pip install wdb.server

In virtualenv or with a different python installation:

    $ pip install wdb

(You must have the server installed and running)

View on GitHub


5.  lptrace

lptrace is strace for Python programs. It lets you see in real-time what functions a Python program is running. It's particularly useful to debug weird issues on production.

For example, let's debug a non-trivial program, the Python SimpleHTTPServer. First, let's run the server:

vagrant@precise32:/vagrant$ python -m SimpleHTTPServer 8080 &
[1] 1818
vagrant@precise32:/vagrant$ Serving HTTP on 0.0.0.0 port 8080 ...

Now let's connect lptrace to it:

vagrant@precise32:/vagrant$ sudo python lptrace -p 1818
...
fileno (/usr/lib/python2.7/SocketServer.py:438)
meth (/usr/lib/python2.7/socket.py:223)

fileno (/usr/lib/python2.7/SocketServer.py:438)
meth (/usr/lib/python2.7/socket.py:223)

_handle_request_noblock (/usr/lib/python2.7/SocketServer.py:271)
get_request (/usr/lib/python2.7/SocketServer.py:446)
accept (/usr/lib/python2.7/socket.py:201)
__init__ (/usr/lib/python2.7/socket.py:185)
verify_request (/usr/lib/python2.7/SocketServer.py:296)
process_request (/usr/lib/python2.7/SocketServer.py:304)
finish_request (/usr/lib/python2.7/SocketServer.py:321)
__init__ (/usr/lib/python2.7/SocketServer.py:632)
setup (/usr/lib/python2.7/SocketServer.py:681)
makefile (/usr/lib/python2.7/socket.py:212)
__init__ (/usr/lib/python2.7/socket.py:246)
makefile (/usr/lib/python2.7/socket.py:212)
__init__ (/usr/lib/python2.7/socket.py:246)
handle (/usr/lib/python2.7/BaseHTTPServer.py:336)
handle_one_request (/usr/lib/python2.7/BaseHTTPServer.py:301)
^CReceived Ctrl-C, quitting
vagrant@precise32:/vagrant$

You can see that the server is handling the request in real time! After pressing Ctrl-C, the trace is removed and the program execution resumes normally.

View on GitHub


6.  python-manhole

Debugging manhole for python applications.

Manhole is in-process service that will accept unix domain socket connections and present the stacktraces for all threads and an interactive prompt. It can either work as a python daemon thread waiting for connections at all times or a signal handler (stopping your application and waiting for a connection).

Access to the socket is restricted to the application's effective user id or root.

This is just like Twisted's manhole. It's simpler (no dependencies), it only runs on Unix domain sockets (in contrast to Twisted's manhole which can run on telnet or ssh) and it integrates well with various types of applications.

Usage

Install it:

pip install manhole

You can put this in your django settings, wsgi app file, some module that's always imported early etc:

import manhole
manhole.install() # this will start the daemon thread

# and now you start your app, eg: server.serve_forever()

Now in a shell you can do either of these:

netcat -U /tmp/manhole-1234
socat - unix-connect:/tmp/manhole-1234
socat readline unix-connect:/tmp/manhole-1234

Socat with readline is best (history, editing etc). If your socat doesn't have readline try this.

Sample output:

$ nc -U /tmp/manhole-1234

Python 2.7.3 (default, Apr 10 2013, 06:20:15)
[GCC 4.6.3] on linux2
Type "help", "copyright", "credits" or "license" for more information.
(InteractiveConsole)
>>> dir()
['__builtins__', 'dump_stacktraces', 'os', 'socket', 'sys', 'traceback']
>>> print 'foobar'
foobar

View on GitHub


7.  Pyringe

Pyringe is a python debugger capable of attaching to running processes, inspecting their state and even of injecting python code into them while they're running. With pyringe, you can list threads, get tracebacks, inspect locals/globals/builtins of running functions, all without having to prepare your program for it.

How do I use it?

You can start the debugger by executing python -m pyringe. Alternatively:

import pyringe
pyringe.interact()

If that reminds you of the code module, good; this is intentional.
After starting the debugger, you'll be greeted by what behaves almost like a regular python REPL.
Try the following:

==> pid:[None] #threads:[0] current thread:[None]
>>> help()
Available commands:
 attach: Attach to the process with the given pid.
 bt: Get a backtrace of the current position.
 [...]
==> pid:[None] #threads:[0] current thread:[None]
>>> attach(12679)
==> pid:[12679] #threads:[11] current thread:[140108099462912]
>>> threads()
[140108099462912, 140108107855616, 140108116248323, 140108124641024, 140108133033728, 140108224739072, 140108233131776, 140108141426432, 140108241524480, 140108249917184, 140108269324032]

The IDs you see here correspond to what threading.current_thread().ident would tell you.
All debugger functions are just regular python functions that have been exposed to the REPL, so you can do things like the following.

==> pid:[12679] #threads:[11] current thread:[140108099462912]
>>> for tid in threads():
...   if not tid % 10:
...     thread(tid)
...     bt()
... 
Traceback (most recent call last):
  File "/usr/lib/python2.7/threading.py", line 524, in __bootstrap
    self.__bootstrap_inner()
  File "/usr/lib/python2.7/threading.py", line 551, in __bootstrap_inner
    self.run()
  File "/usr/lib/python2.7/threading.py", line 504, in run
    self.__target(*self.__args, **self.__kwargs)
  File "./test.py", line 46, in Idle
    Thread_2_Func(1)
  File "./test.py", line 40, in Wait
    time.sleep(n)
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> 

You can access the inferior's locals and inspect them like so:

==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> inflocals()
{'a': <proxy of A object at remote 0x1d9b290>, 'LOL': 'success!', 'b': <proxy of B object at remote 0x1d988c0>, 'n': 1}
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> p('a')
<proxy of A object at remote 0x1d9b290>
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> p('a').attr
'Some_magic_string'
==> pid:[12679] #threads:[11] current thread:[140108241524480]
>>> 

And sure enough, the definition of a's class reads:

class Example(object):
  cl_attr = False
  def __init__(self):
    self.attr = 'Some_magic_string'

There's limits to how far this proxying of objects goes, and everything that isn't trivial data will show up as strings (like '<function at remote 0x1d957d0>').

View on GitHub


8.  python-hunter

Hunter is a flexible code tracing toolkit, not for measuring coverage, but for debugging, logging, inspection and other nefarious purposes. It has a simple Python API, a convenient terminal API and a CLI tool to attach to processes.

Installation

pip install hunter

Documentation

https://python-hunter.readthedocs.io/

Getting started

Basic use involves passing various filters to the trace option. An example:

import hunter
hunter.trace(module='posixpath', action=hunter.CallPrinter)

import os
os.path.join('a', 'b')

That would result in:

>>> os.path.join('a', 'b')
         /usr/lib/python3.6/posixpath.py:75    call      => join(a='a')
         /usr/lib/python3.6/posixpath.py:80    line         a = os.fspath(a)
         /usr/lib/python3.6/posixpath.py:81    line         sep = _get_sep(a)
         /usr/lib/python3.6/posixpath.py:41    call         => _get_sep(path='a')
         /usr/lib/python3.6/posixpath.py:42    line            if isinstance(path, bytes):
         /usr/lib/python3.6/posixpath.py:45    line            return '/'
         /usr/lib/python3.6/posixpath.py:45    return       <= _get_sep: '/'
         /usr/lib/python3.6/posixpath.py:82    line         path = a
         /usr/lib/python3.6/posixpath.py:83    line         try:
         /usr/lib/python3.6/posixpath.py:84    line         if not p:
         /usr/lib/python3.6/posixpath.py:86    line         for b in map(os.fspath, p):
         /usr/lib/python3.6/posixpath.py:87    line         if b.startswith(sep):
         /usr/lib/python3.6/posixpath.py:89    line         elif not path or path.endswith(sep):
         /usr/lib/python3.6/posixpath.py:92    line         path += sep + b
         /usr/lib/python3.6/posixpath.py:86    line         for b in map(os.fspath, p):
         /usr/lib/python3.6/posixpath.py:96    line         return path
         /usr/lib/python3.6/posixpath.py:96    return    <= join: 'a/b'
'a/b'

In a terminal it would look like:

https://raw.githubusercontent.com/ionelmc/python-hunter/master/docs/code-trace.png

Another useful scenario is to ignore all standard modules and force colors to make them stay even if the output is redirected to a file.

import hunter
hunter.trace(stdlib=False, action=hunter.CallPrinter(force_colors=True))

View on GitHub


9.  line_profiler

line_profiler is a module for doing line-by-line profiling of functions. kernprof is a convenient script for running either line_profiler or the Python standard library's cProfile or profile modules, depending on what is available.

Installation

Note: As of version 2.1.2, pip install line_profiler does not work. Please install as follows until it is fixed in the next release:

git clone https://github.com/rkern/line_profiler.git
find line_profiler -name '*.pyx' -exec cython {} \;
cd line_profiler
pip install . --user

Releases of line_profiler can be installed using pip:

$ pip install line_profiler

Source releases and any binaries can be downloaded from the PyPI link.

http://pypi.python.org/pypi/line_profiler

To check out the development sources, you can use Git:

$ git clone https://github.com/rkern/line_profiler.git

You may also download source tarballs of any snapshot from that URL.

Source releases will require a C compiler in order to build line_profiler. In addition, git checkouts will also require Cython >= 0.10. Source releases on PyPI should contain the pregenerated C sources, so Cython should not be required in that case.

kernprof is a single-file pure Python script and does not require a compiler. If you wish to use it to run cProfile and not line-by-line profiling, you may copy it to a directory on your PATH manually and avoid trying to build any C extensions.

View on GitHub


10.  Memory Profiler

This is a python module for monitoring memory consumption of a process as well as line-by-line analysis of memory consumption for python programs. It is a pure python module which depends on the psutil module.

Installation

To install through easy_install or pip:

$ easy_install -U memory_profiler # pip install -U memory_profiler

To install from source, download the package, extract and type:

$ python setup.py install

Usage

line-by-line memory usage

The line-by-line memory usage mode is used much in the same way of the line_profiler: first decorate the function you would like to profile with @profile and then run the script with a special script (in this case with specific arguments to the Python interpreter).

In the following example, we create a simple function my_func that allocates lists a, b and then deletes b:

@profile
def my_func():
    a = [1] * (10 ** 6)
    b = [2] * (2 * 10 ** 7)
    del b
    return a

if __name__ == '__main__':
    my_func()

Execute the code passing the option -m memory_profiler to the python interpreter to load the memory_profiler module and print to stdout the line-by-line analysis. If the file name was example.py, this would result in:

$ python -m memory_profiler example.py

Output will follow:

Line #    Mem usage  Increment   Line Contents
==============================================
     3                           @profile
     4      5.97 MB    0.00 MB   def my_func():
     5     13.61 MB    7.64 MB       a = [1] * (10 ** 6)
     6    166.20 MB  152.59 MB       b = [2] * (2 * 10 ** 7)
     7     13.61 MB -152.59 MB       del b
     8     13.61 MB    0.00 MB       return a

The first column represents the line number of the code that has been profiled, the second column (Mem usage) the memory usage of the Python interpreter after that line has been executed. The third column (Increment) represents the difference in memory of the current line with respect to the last one. The last column (Line Contents) prints the code that has been profiled.

View on GitHub


11.  py-spy

py-spy is a sampling profiler for Python programs. It lets you visualize what your Python program is spending time on without restarting the program or modifying the code in any way. py-spy is extremely low overhead: it is written in Rust for speed and doesn't run in the same process as the profiled Python program. This means py-spy is safe to use against production Python code.

py-spy works on Linux, OSX, Windows and FreeBSD, and supports profiling all recent versions of the CPython interpreter (versions 2.3-2.7 and 3.3-3.10).

Installation

Prebuilt binary wheels can be installed from PyPI with:

pip install py-spy

You can also download prebuilt binaries from the GitHub Releases Page.

If you're a Rust user, py-spy can also be installed with: cargo install py-spy.

On macOS, py-spy is in Homebrew and can be installed with brew install py-spy.

On Arch Linux, py-spy is in AUR and can be installed with yay -S py-spy.

On Alpine Linux, py-spy is in testing repository and can be installed with apk add py-spy --update-cache --repository http://dl-3.alpinelinux.org/alpine/edge/testing/ --allow-untrusted.

Usage

py-spy works from the command line and takes either the PID of the program you want to sample from or the command line of the python program you want to run. py-spy has three subcommands record, top and dump:

record

py-spy supports recording profiles to a file using the record command. For example, you can generate a flame graph of your python process by going:

py-spy record -o profile.svg --pid 12345
# OR
py-spy record -o profile.svg -- python myprogram.py

View on GitHub


12.  Pyflame

Pyflame is a high performance profiling tool that generates flame graphs for Python. Pyflame is implemented in C++, and uses the Linux ptrace(2) system call to collect profiling information. It can take snapshots of the Python call stack without explicit instrumentation, meaning you can profile a program without modifying its source code. Pyflame is capable of profiling embedded Python interpreters like uWSGI. It fully supports profiling multi-threaded Python programs.

Pyflame usually introduces significantly less overhead than the builtin profile (or cProfile) modules, and emits richer profiling data. The profiling overhead is low enough that you can use it to profile live processes in production.

Quickstart

Building And Installing

For Debian/Ubuntu, install the following:

# Install build dependencies on Debian or Ubuntu.
sudo apt-get install autoconf automake autotools-dev g++ pkg-config python-dev python3-dev libtool make

Once you have the build dependencies installed:

./autogen.sh
./configure
make

The make command will produce an executable at src/pyflame that you can run and use.

Optionally, if you have virtualenv installed, you can test the executable you produced using make check.

Using Pyflame

The full documentation for using Pyflame is here. But here's a quick guide:

# Attach to PID 12345 and profile it for 1 second
pyflame -p 12345

# Attach to PID 768 and profile it for 5 seconds, sampling every 0.01 seconds
pyflame -s 5 -r 0.01 -p 768

# Run py.test against tests/, emitting sample data to prof.txt
pyflame -o prof.txt -t py.test tests/

In all of these cases you will get flame graph data on stdout (or to a file if you used -o). This data is in the format expected by flamegraph.pl, which you can find here.

View on GitHub


13.  vprof

vprof is a Python package providing rich and interactive visualizations for various Python program characteristics such as running time and memory usage. It supports Python 3.4+ and distributed under BSD license.

The project is in active development and some of its features might not work as expected.

Installation

vprof can be installed from PyPI

pip install vprof

To build vprof from sources, clone this repository and execute

python3 setup.py deps_install && python3 setup.py build_ui && python3 setup.py install

To install just vprof dependencies, run

python3 setup.py deps_install

Usage

vprof -c <config> <src>

<config> is a combination of supported modes:

  • c - CPU flame graph ⚠️ Not available for windows #62

Shows CPU flame graph for <src>.

  • p - profiler

Runs built-in Python profiler on <src> and displays results.

  • m - memory graph

Shows objects that are tracked by CPython GC and left in memory after code execution. Also shows process memory usage after execution of each line of <src>.

  • h - code heatmap

Displays all executed code of <src> with line run times and execution counts.

View on GitHub


14.  Django Debug Toolbar

The Django Debug Toolbar is a configurable set of panels that display various debug information about the current request/response and when clicked, display more details about the panel's content.


Here's a screenshot of the toolbar in action:

Django Debug Toolbar screenshot

In addition to the built-in panels, a number of third-party panels are contributed by the community.

The current stable version of the Debug Toolbar is 3.6.0. It works on Django ≥ 3.2.4.

View on GitHub


15.  django-devserver

A drop in replacement for Django's built-in runserver command. Features include:

  • An extendable interface for handling things such as real-time logging.
  • Integration with the werkzeug interactive debugger.
  • Threaded (default) and multi-process development servers.
  • Ability to specify a WSGI application as your target environment.

Note

django-devserver works on Django 1.3 and newer

Installation

To install the latest stable version:

pip install git+git://github.com/dcramer/django-devserver#egg=django-devserver

django-devserver has some optional dependancies, which we highly recommend installing.

  • pip install sqlparse -- pretty SQL formatting
  • pip install werkzeug -- interactive debugger
  • pip install guppy -- tracks memory usage (required for MemoryUseModule)
  • pip install line_profiler -- does line-by-line profiling (required for LineProfilerModule)

You will need to include devserver in your INSTALLED_APPS:

INSTALLED_APPS = (
    ...
    'devserver',
)

If you're using django.contrib.staticfiles or any other apps with management command runserver, make sure to put devserver above any of them (or below, for Django<1.7). Otherwise devserver will log an error, but it will fail to work properly.

View on GitHub


16.  Flask Debug-toolbar

This is a port of the excellent django-debug-toolbar for Flask applications.

Installation

Installing is simple with pip:

$ pip install flask-debugtoolbar

Usage

Setting up the debug toolbar is simple:

from flask import Flask
from flask_debugtoolbar import DebugToolbarExtension

app = Flask(__name__)

# the toolbar is only enabled in debug mode:
app.debug = True

# set a 'SECRET_KEY' to enable the Flask session cookies
app.config['SECRET_KEY'] = '<replace with a secret key>'

toolbar = DebugToolbarExtension(app)

The toolbar will automatically be injected into Jinja templates when debug mode is on. In production, setting app.debug = False will disable the toolbar.

View on GitHub


17.  IceCream

Do you ever use print() or log() to debug your code? Of course you do. IceCream, or ic for short, makes print debugging a little sweeter.

ic() is like print(), but better:

  1. It prints both expressions/variable names and their values.
  2. It's 40% faster to type.
  3. Data structures are pretty printed.
  4. Output is syntax highlighted.
  5. It optionally includes program context: filename, line number, and parent function.

IceCream is well tested, permissively licensed, and supports Python 2, Python 3, PyPy2, and PyPy3. (Python 3.11 support is forthcoming.)

Inspect Variables

Have you ever printed variables or expressions to debug your program? If you've ever typed something like

print(foo('123'))

or the more thorough

print("foo('123')", foo('123'))

then ic() will put a smile on your face. With arguments, ic() inspects itself and prints both its own arguments and the values of those arguments.

from icecream import ic

def foo(i):
    return i + 333

ic(foo(123))

Prints

ic| foo(123): 456

Similarly,

d = {'key': {1: 'one'}}
ic(d['key'][1])

class klass():
    attr = 'yep'
ic(klass.attr)

Prints

ic| d['key'][1]: 'one'
ic| klass.attr: 'yep'

Just give ic() a variable or expression and you're done. Easy.

View on GitHub


18.  pyelftools

pyelftools is a pure-Python library for parsing and analyzing ELF files and DWARF debugging information. See the User's guide for more details.

Pre-requisites

As a user of pyelftools, one only needs Python 3 to run. For hacking on pyelftools the requirements are a bit more strict, please see the hacking guide.

Installing

pyelftools can be installed from PyPI (Python package index):

> pip install pyelftools

Alternatively, you can download the source distribution for the most recent and historic versions from the Downloads tab on the pyelftools project page (by going to Tags). Then, you can install from source, as usual:

> python setup.py install

Since pyelftools is a work in progress, it's recommended to have the most recent version of the code. This can be done by downloading the master zip file or just cloning the Git repository.

Since pyelftools has no external dependencies, it's also easy to use it without installing, by locally adjusting PYTHONPATH.

View on GitHub


FAQ about Debugging Tools python

  • How many types of debugging are in Python?

Debugging in any programming language typically involves two types of errors: syntax or logical. Syntax errors are those where the programming language commands are not interpreted by the compiler or interpreter because of a problem with how the program is written.

  • Best Debugging Tools include:

Chrome DevTools, Progress Telerik Fiddler, GDB (GNU Debugger), Data Display Debugger, SonarLint, Froglogic Squish, and TotalView HPC Debugging Software.

  • Why is it called debugging?

The terms "bug" and "debugging" are popularly attributed to Admiral Grace Hopper in the 1940s. While she was working on a Mark II computer at Harvard University, her associates discovered a moth stuck in a relay and thereby impeding operation, whereupon she remarked that they were "debugging" the system.

  • Why do we need debugging?

Debugging is important because it allows software engineers and developers to fix errors in a program before releasing it to the public. It's a complementary process to testing, which involves learning how an error affects a program overall.


Related videos:

Python Tutorial - Introduction to DEBUGGING


Related posts:

#python 

Tyrique  Littel

Tyrique Littel

1600135200

How to Install OpenJDK 11 on CentOS 8

What is OpenJDK?

OpenJDk or Open Java Development Kit is a free, open-source framework of the Java Platform, Standard Edition (or Java SE). It contains the virtual machine, the Java Class Library, and the Java compiler. The difference between the Oracle OpenJDK and Oracle JDK is that OpenJDK is a source code reference point for the open-source model. Simultaneously, the Oracle JDK is a continuation or advanced model of the OpenJDK, which is not open source and requires a license to use.

In this article, we will be installing OpenJDK on Centos 8.

#tutorials #alternatives #centos #centos 8 #configuration #dnf #frameworks #java #java development kit #java ee #java environment variables #java framework #java jdk #java jre #java platform #java sdk #java se #jdk #jre #open java development kit #open source #openjdk #openjdk 11 #openjdk 8 #openjdk runtime environment

Joseph  Murray

Joseph Murray

1623302550

Top 5 Java Web Application Technologies You Should Master in 2021

Web Development in Java

Java is a commonly used language for web development, especially on the server-side. Java web applications are distributed applications that run on the internet. Web development with Java allows us to create dynamic web pages where users can interact with the interface.

There are various ways through which you can create dynamic web pages in Java. The Java EE (Enterprise Edition) platform provides various Java technologies for web development to developers. Services like distributed computing, web services, etc. are provided by Java EE. Applications can be developed in Java without using any additional scripting language. Let us see how web applications are made via Java.

**Java Web Application **

Java Web Application Technologies

#software development #java #java web applications #web applications #java web application technologies #top 5 java web application technologies you should master

Daniel  Hughes

Daniel Hughes

1649214000

LaravelS: Glue for using Swoole in Laravel Or Lumen.

🚀 LaravelS is an out-of-the-box adapter between Swoole and Laravel/Lumen.

Please Watch this repository to get the latest updates.

中文文档

Table of Contents

Features

Built-in Http/WebSocket server

Multi-port mixed protocol

Custom process

Memory resident

Asynchronous event listening

Asynchronous task queue

Millisecond cron job

Common Components

Gracefully reload

Automatically reload after modifying code

Support Laravel/Lumen both, good compatibility

Simple & Out of the box

Benchmark

Which is the fastest web framework?

TechEmpower Framework Benchmarks

Requirements

DependencyRequirement
PHP>= 5.5.9 Recommend PHP7+
Swoole>= 1.7.19 No longer support PHP5 since 2.0.12 Recommend 4.5.0+
Laravel/Lumen>= 5.1 Recommend 8.0+

Install

1.Require package via Composer(packagist).

composer require "hhxsv5/laravel-s:~3.7.0" -vvv
# Make sure that your composer.lock file is under the VCS

2.Register service provider(pick one of two).

Laravel: in config/app.php file, Laravel 5.5+ supports package discovery automatically, you should skip this step

'providers' => [
    //...
    Hhxsv5\LaravelS\Illuminate\LaravelSServiceProvider::class,
],

Lumen: in bootstrap/app.php file

$app->register(Hhxsv5\LaravelS\Illuminate\LaravelSServiceProvider::class);

3.Publish configuration and binaries.

After upgrading LaravelS, you need to republish; click here to see the change notes of each version.

php artisan laravels publish
# Configuration: config/laravels.php
# Binary: bin/laravels bin/fswatch bin/inotify

4.Change config/laravels.php: listen_ip, listen_port, refer Settings.

5.Performance tuning

Adjust kernel parameters

Number of Workers: LaravelS uses Swoole's Synchronous IO mode, the larger the worker_num setting, the better the concurrency performance, but it will cause more memory usage and process switching overhead. If one request takes 100ms, in order to provide 1000QPS concurrency, at least 100 Worker processes need to be configured. The calculation method is: worker_num = 1000QPS/(1s/1ms) = 100, so incremental pressure testing is needed to calculate the best worker_num.

Number of Task Workers

Run

Please read the notices carefully before running, Important notices(IMPORTANT).

  • Commands: php bin/laravels {start|stop|restart|reload|info|help}.
CommandDescription
startStart LaravelS, list the processes by "ps -ef|grep laravels"
stopStop LaravelS, and trigger the method onStop of Custom process
restartRestart LaravelS: Stop gracefully before starting; The service is unavailable until startup is complete
reloadReload all Task/Worker/Timer processes which contain your business codes, and trigger the method onReload of Custom process, CANNOT reload Master/Manger processes. After modifying config/laravels.php, you only have to call restart to restart
infoDisplay component version information
helpDisplay help information
  • Boot options for the commands start and restart.
OptionDescription
-d|--daemonizeRun as a daemon, this option will override the swoole.daemonize setting in laravels.php
-e|--envThe environment the command should run under, such as --env=testing will use the configuration file .env.testing firstly, this feature requires Laravel 5.2+
-i|--ignoreIgnore checking PID file of Master process
-x|--x-versionThe version(branch) of the current project, stored in $_ENV/$_SERVER, access via $_ENV['X_VERSION'] $_SERVER['X_VERSION'] $request->server->get('X_VERSION')
  • Runtime files: start will automatically execute php artisan laravels config and generate these files, developers generally don't need to pay attention to them, it's recommended to add them to .gitignore.
FileDescription
storage/laravels.confLaravelS's runtime configuration file
storage/laravels.pidPID file of Master process
storage/laravels-timer-process.pidPID file of the Timer process
storage/laravels-custom-processes.pidPID file of all custom processes

Deploy

It is recommended to supervise the main process through Supervisord, the premise is without option -d and to set swoole.daemonize to false.

[program:laravel-s-test]
directory=/var/www/laravel-s-test
command=/usr/local/bin/php bin/laravels start -i
numprocs=1
autostart=true
autorestart=true
startretries=3
user=www-data
redirect_stderr=true
stdout_logfile=/var/log/supervisor/%(program_name)s.log

Cooperate with Nginx (Recommended)

Demo.

gzip on;
gzip_min_length 1024;
gzip_comp_level 2;
gzip_types text/plain text/css text/javascript application/json application/javascript application/x-javascript application/xml application/x-httpd-php image/jpeg image/gif image/png font/ttf font/otf image/svg+xml;
gzip_vary on;
gzip_disable "msie6";
upstream swoole {
    # Connect IP:Port
    server 127.0.0.1:5200 weight=5 max_fails=3 fail_timeout=30s;
    # Connect UnixSocket Stream file, tips: put the socket file in the /dev/shm directory to get better performance
    #server unix:/yourpath/laravel-s-test/storage/laravels.sock weight=5 max_fails=3 fail_timeout=30s;
    #server 192.168.1.1:5200 weight=3 max_fails=3 fail_timeout=30s;
    #server 192.168.1.2:5200 backup;
    keepalive 16;
}
server {
    listen 80;
    # Don't forget to bind the host
    server_name laravels.com;
    root /yourpath/laravel-s-test/public;
    access_log /yourpath/log/nginx/$server_name.access.log  main;
    autoindex off;
    index index.html index.htm;
    # Nginx handles the static resources(recommend enabling gzip), LaravelS handles the dynamic resource.
    location / {
        try_files $uri @laravels;
    }
    # Response 404 directly when request the PHP file, to avoid exposing public/*.php
    #location ~* \.php$ {
    #    return 404;
    #}
    location @laravels {
        # proxy_connect_timeout 60s;
        # proxy_send_timeout 60s;
        # proxy_read_timeout 120s;
        proxy_http_version 1.1;
        proxy_set_header Connection "";
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Real-PORT $remote_port;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header Host $http_host;
        proxy_set_header Scheme $scheme;
        proxy_set_header Server-Protocol $server_protocol;
        proxy_set_header Server-Name $server_name;
        proxy_set_header Server-Addr $server_addr;
        proxy_set_header Server-Port $server_port;
        # "swoole" is the upstream
        proxy_pass http://swoole;
    }
}

Cooperate with Apache

LoadModule proxy_module /yourpath/modules/mod_proxy.so
LoadModule proxy_balancer_module /yourpath/modules/mod_proxy_balancer.so
LoadModule lbmethod_byrequests_module /yourpath/modules/mod_lbmethod_byrequests.so
LoadModule proxy_http_module /yourpath/modules/mod_proxy_http.so
LoadModule slotmem_shm_module /yourpath/modules/mod_slotmem_shm.so
LoadModule rewrite_module /yourpath/modules/mod_rewrite.so
LoadModule remoteip_module /yourpath/modules/mod_remoteip.so
LoadModule deflate_module /yourpath/modules/mod_deflate.so

<IfModule deflate_module>
    SetOutputFilter DEFLATE
    DeflateCompressionLevel 2
    AddOutputFilterByType DEFLATE text/html text/plain text/css text/javascript application/json application/javascript application/x-javascript application/xml application/x-httpd-php image/jpeg image/gif image/png font/ttf font/otf image/svg+xml
</IfModule>

<VirtualHost *:80>
    # Don't forget to bind the host
    ServerName www.laravels.com
    ServerAdmin hhxsv5@sina.com

    DocumentRoot /yourpath/laravel-s-test/public;
    DirectoryIndex index.html index.htm
    <Directory "/">
        AllowOverride None
        Require all granted
    </Directory>

    RemoteIPHeader X-Forwarded-For

    ProxyRequests Off
    ProxyPreserveHost On
    <Proxy balancer://laravels>  
        BalancerMember http://192.168.1.1:5200 loadfactor=7
        #BalancerMember http://192.168.1.2:5200 loadfactor=3
        #BalancerMember http://192.168.1.3:5200 loadfactor=1 status=+H
        ProxySet lbmethod=byrequests
    </Proxy>
    #ProxyPass / balancer://laravels/
    #ProxyPassReverse / balancer://laravels/

    # Apache handles the static resources, LaravelS handles the dynamic resource.
    RewriteEngine On
    RewriteCond %{DOCUMENT_ROOT}%{REQUEST_FILENAME} !-d
    RewriteCond %{DOCUMENT_ROOT}%{REQUEST_FILENAME} !-f
    RewriteRule ^/(.*)$ balancer://laravels%{REQUEST_URI} [P,L]

    ErrorLog ${APACHE_LOG_DIR}/www.laravels.com.error.log
    CustomLog ${APACHE_LOG_DIR}/www.laravels.com.access.log combined
</VirtualHost>

Enable WebSocket server

The Listening address of WebSocket Sever is the same as Http Server.

1.Create WebSocket Handler class, and implement interface WebSocketHandlerInterface.The instant is automatically instantiated when start, you do not need to manually create it.

namespace App\Services;
use Hhxsv5\LaravelS\Swoole\WebSocketHandlerInterface;
use Swoole\Http\Request;
use Swoole\Http\Response;
use Swoole\WebSocket\Frame;
use Swoole\WebSocket\Server;
/**
 * @see https://www.swoole.co.uk/docs/modules/swoole-websocket-server
 */
class WebSocketService implements WebSocketHandlerInterface
{
    // Declare constructor without parameters
    public function __construct()
    {
    }
    // public function onHandShake(Request $request, Response $response)
    // {
           // Custom handshake: https://www.swoole.co.uk/docs/modules/swoole-websocket-server-on-handshake
           // The onOpen event will be triggered automatically after a successful handshake
    // }
    public function onOpen(Server $server, Request $request)
    {
        // Before the onOpen event is triggered, the HTTP request to establish the WebSocket has passed the Laravel route,
        // so Laravel's Request, Auth information are readable, Session is readable and writable, but only in the onOpen event.
        // \Log::info('New WebSocket connection', [$request->fd, request()->all(), session()->getId(), session('xxx'), session(['yyy' => time()])]);
        // The exceptions thrown here will be caught by the upper layer and recorded in the Swoole log. Developers need to try/catch manually.
        $server->push($request->fd, 'Welcome to LaravelS');
    }
    public function onMessage(Server $server, Frame $frame)
    {
        // \Log::info('Received message', [$frame->fd, $frame->data, $frame->opcode, $frame->finish]);
        // The exceptions thrown here will be caught by the upper layer and recorded in the Swoole log. Developers need to try/catch manually.
        $server->push($frame->fd, date('Y-m-d H:i:s'));
    }
    public function onClose(Server $server, $fd, $reactorId)
    {
        // The exceptions thrown here will be caught by the upper layer and recorded in the Swoole log. Developers need to try/catch manually.
    }
}

2.Modify config/laravels.php.

// ...
'websocket'      => [
    'enable'  => true, // Note: set enable to true
    'handler' => \App\Services\WebSocketService::class,
],
'swoole'         => [
    //...
    // Must set dispatch_mode in (2, 4, 5), see https://www.swoole.co.uk/docs/modules/swoole-server/configuration
    'dispatch_mode' => 2,
    //...
],
// ...

3.Use SwooleTable to bind FD & UserId, optional, Swoole Table Demo. Also you can use the other global storage services, like Redis/Memcached/MySQL, but be careful that FD will be possible conflicting between multiple Swoole Servers.

4.Cooperate with Nginx (Recommended)

Refer WebSocket Proxy

map $http_upgrade $connection_upgrade {
    default upgrade;
    ''      close;
}
upstream swoole {
    # Connect IP:Port
    server 127.0.0.1:5200 weight=5 max_fails=3 fail_timeout=30s;
    # Connect UnixSocket Stream file, tips: put the socket file in the /dev/shm directory to get better performance
    #server unix:/yourpath/laravel-s-test/storage/laravels.sock weight=5 max_fails=3 fail_timeout=30s;
    #server 192.168.1.1:5200 weight=3 max_fails=3 fail_timeout=30s;
    #server 192.168.1.2:5200 backup;
    keepalive 16;
}
server {
    listen 80;
    # Don't forget to bind the host
    server_name laravels.com;
    root /yourpath/laravel-s-test/public;
    access_log /yourpath/log/nginx/$server_name.access.log  main;
    autoindex off;
    index index.html index.htm;
    # Nginx handles the static resources(recommend enabling gzip), LaravelS handles the dynamic resource.
    location / {
        try_files $uri @laravels;
    }
    # Response 404 directly when request the PHP file, to avoid exposing public/*.php
    #location ~* \.php$ {
    #    return 404;
    #}
    # Http and WebSocket are concomitant, Nginx identifies them by "location"
    # !!! The location of WebSocket is "/ws"
    # Javascript: var ws = new WebSocket("ws://laravels.com/ws");
    location =/ws {
        # proxy_connect_timeout 60s;
        # proxy_send_timeout 60s;
        # proxy_read_timeout: Nginx will close the connection if the proxied server does not send data to Nginx in 60 seconds; At the same time, this close behavior is also affected by heartbeat setting of Swoole.
        # proxy_read_timeout 60s;
        proxy_http_version 1.1;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Real-PORT $remote_port;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header Host $http_host;
        proxy_set_header Scheme $scheme;
        proxy_set_header Server-Protocol $server_protocol;
        proxy_set_header Server-Name $server_name;
        proxy_set_header Server-Addr $server_addr;
        proxy_set_header Server-Port $server_port;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection $connection_upgrade;
        proxy_pass http://swoole;
    }
    location @laravels {
        # proxy_connect_timeout 60s;
        # proxy_send_timeout 60s;
        # proxy_read_timeout 60s;
        proxy_http_version 1.1;
        proxy_set_header Connection "";
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Real-PORT $remote_port;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header Host $http_host;
        proxy_set_header Scheme $scheme;
        proxy_set_header Server-Protocol $server_protocol;
        proxy_set_header Server-Name $server_name;
        proxy_set_header Server-Addr $server_addr;
        proxy_set_header Server-Port $server_port;
        proxy_pass http://swoole;
    }
}

5.Heartbeat setting

Heartbeat setting of Swoole

// config/laravels.php
'swoole' => [
    //...
    // All connections are traversed every 60 seconds. If a connection does not send any data to the server within 600 seconds, the connection will be forced to close.
    'heartbeat_idle_time'      => 600,
    'heartbeat_check_interval' => 60,
    //...
],

Proxy read timeout of Nginx

# Nginx will close the connection if the proxied server does not send data to Nginx in 60 seconds
proxy_read_timeout 60s;

6.Push data in controller

namespace App\Http\Controllers;
class TestController extends Controller
{
    public function push()
    {
        $fd = 1; // Find fd by userId from a map [userId=>fd].
        /**@var \Swoole\WebSocket\Server $swoole */
        $swoole = app('swoole');
        $success = $swoole->push($fd, 'Push data to fd#1 in Controller');
        var_dump($success);
    }
}

Listen events

System events

Usually, you can reset/destroy some global/static variables, or change the current Request/Response object.

laravels.received_request After LaravelS parsed Swoole\Http\Request to Illuminate\Http\Request, before Laravel's Kernel handles this request.

// Edit file `app/Providers/EventServiceProvider.php`, add the following code into method `boot`
// If no variable $events, you can also call Facade \Event::listen(). 
$events->listen('laravels.received_request', function (\Illuminate\Http\Request $req, $app) {
    $req->query->set('get_key', 'hhxsv5');// Change query of request
    $req->request->set('post_key', 'hhxsv5'); // Change post of request
});

laravels.generated_response After Laravel's Kernel handled the request, before LaravelS parses Illuminate\Http\Response to Swoole\Http\Response.

// Edit file `app/Providers/EventServiceProvider.php`, add the following code into method `boot`
// If no variable $events, you can also call Facade \Event::listen(). 
$events->listen('laravels.generated_response', function (\Illuminate\Http\Request $req, \Symfony\Component\HttpFoundation\Response $rsp, $app) {
    $rsp->headers->set('header-key', 'hhxsv5');// Change header of response
});

Customized asynchronous events

This feature depends on AsyncTask of Swoole, your need to set swoole.task_worker_num in config/laravels.php firstly. The performance of asynchronous event processing is influenced by number of Swoole task process, you need to set task_worker_num appropriately.

1.Create event class.

use Hhxsv5\LaravelS\Swoole\Task\Event;
class TestEvent extends Event
{
    protected $listeners = [
        // Listener list
        TestListener1::class,
        // TestListener2::class,
    ];
    private $data;
    public function __construct($data)
    {
        $this->data = $data;
    }
    public function getData()
    {
        return $this->data;
    }
}

2.Create listener class.

use Hhxsv5\LaravelS\Swoole\Task\Task;
use Hhxsv5\LaravelS\Swoole\Task\Listener;
class TestListener1 extends Listener
{
    /**
     * @var TestEvent
     */
    protected $event;
    
    public function handle()
    {
        \Log::info(__CLASS__ . ':handle start', [$this->event->getData()]);
        sleep(2);// Simulate the slow codes
        // Deliver task in CronJob, but NOT support callback finish() of task.
        // Note: Modify task_ipc_mode to 1 or 2 in config/laravels.php, see https://www.swoole.co.uk/docs/modules/swoole-server/configuration
        $ret = Task::deliver(new TestTask('task data'));
        var_dump($ret);
        // The exceptions thrown here will be caught by the upper layer and recorded in the Swoole log. Developers need to try/catch manually.
    }
}

3.Fire event.

// Create instance of event and fire it, "fire" is asynchronous.
use Hhxsv5\LaravelS\Swoole\Task\Event;
$event = new TestEvent('event data');
// $event->delay(10); // Delay 10 seconds to fire event
// $event->setTries(3); // When an error occurs, try 3 times in total
$success = Event::fire($event);
var_dump($success);// Return true if sucess, otherwise false

Asynchronous task queue

This feature depends on AsyncTask of Swoole, your need to set swoole.task_worker_num in config/laravels.php firstly. The performance of task processing is influenced by number of Swoole task process, you need to set task_worker_num appropriately.

1.Create task class.

use Hhxsv5\LaravelS\Swoole\Task\Task;
class TestTask extends Task
{
    private $data;
    private $result;
    public function __construct($data)
    {
        $this->data = $data;
    }
    // The logic of task handling, run in task process, CAN NOT deliver task
    public function handle()
    {
        \Log::info(__CLASS__ . ':handle start', [$this->data]);
        sleep(2);// Simulate the slow codes
        // The exceptions thrown here will be caught by the upper layer and recorded in the Swoole log. Developers need to try/catch manually.
        $this->result = 'the result of ' . $this->data;
    }
    // Optional, finish event, the logic of after task handling, run in worker process, CAN deliver task 
    public function finish()
    {
        \Log::info(__CLASS__ . ':finish start', [$this->result]);
        Task::deliver(new TestTask2('task2 data')); // Deliver the other task
    }
}

2.Deliver task.

// Create instance of TestTask and deliver it, "deliver" is asynchronous.
use Hhxsv5\LaravelS\Swoole\Task\Task;
$task = new TestTask('task data');
// $task->delay(3);// delay 3 seconds to deliver task
// $task->setTries(3); // When an error occurs, try 3 times in total
$ret = Task::deliver($task);
var_dump($ret);// Return true if sucess, otherwise false

Millisecond cron job

Wrapper cron job base on Swoole's Millisecond Timer, replace Linux Crontab.

1.Create cron job class.

namespace App\Jobs\Timer;
use App\Tasks\TestTask;
use Swoole\Coroutine;
use Hhxsv5\LaravelS\Swoole\Task\Task;
use Hhxsv5\LaravelS\Swoole\Timer\CronJob;
class TestCronJob extends CronJob
{
    protected $i = 0;
    // !!! The `interval` and `isImmediate` of cron job can be configured in two ways(pick one of two): one is to overload the corresponding method, and the other is to pass parameters when registering cron job.
    // --- Override the corresponding method to return the configuration: begin
    public function interval()
    {
        return 1000;// Run every 1000ms
    }
    public function isImmediate()
    {
        return false;// Whether to trigger `run` immediately after setting up
    }
    // --- Override the corresponding method to return the configuration: end
    public function run()
    {
        \Log::info(__METHOD__, ['start', $this->i, microtime(true)]);
        // do something
        // sleep(1); // Swoole < 2.1
        Coroutine::sleep(1); // Swoole>=2.1 Coroutine will be automatically created for run().
        $this->i++;
        \Log::info(__METHOD__, ['end', $this->i, microtime(true)]);

        if ($this->i >= 10) { // Run 10 times only
            \Log::info(__METHOD__, ['stop', $this->i, microtime(true)]);
            $this->stop(); // Stop this cron job, but it will run again after restart/reload.
            // Deliver task in CronJob, but NOT support callback finish() of task.
            // Note: Modify task_ipc_mode to 1 or 2 in config/laravels.php, see https://www.swoole.co.uk/docs/modules/swoole-server/configuration
            $ret = Task::deliver(new TestTask('task data'));
            var_dump($ret);
        }
        // The exceptions thrown here will be caught by the upper layer and recorded in the Swoole log. Developers need to try/catch manually.
    }
}

2.Register cron job.

// Register cron jobs in file "config/laravels.php"
[
    // ...
    'timer'          => [
        'enable' => true, // Enable Timer
        'jobs'   => [ // The list of cron job
            // Enable LaravelScheduleJob to run `php artisan schedule:run` every 1 minute, replace Linux Crontab
            // \Hhxsv5\LaravelS\Illuminate\LaravelScheduleJob::class,
            // Two ways to configure parameters:
            // [\App\Jobs\Timer\TestCronJob::class, [1000, true]], // Pass in parameters when registering
            \App\Jobs\Timer\TestCronJob::class, // Override the corresponding method to return the configuration
        ],
        'max_wait_time' => 5, // Max waiting time of reloading
        // Enable the global lock to ensure that only one instance starts the timer when deploying multiple instances. This feature depends on Redis, please see https://laravel.com/docs/7.x/redis
        'global_lock'     => false,
        'global_lock_key' => config('app.name', 'Laravel'),
    ],
    // ...
];

3.Note: it will launch multiple timers when build the server cluster, so you need to make sure that launch one timer only to avoid running repetitive task.

4.LaravelS v3.4.0 starts to support the hot restart [Reload] Timer process. After LaravelS receives the SIGUSR1 signal, it waits for max_wait_time(default 5) seconds to end the process, then the Manager process will pull up the Timer process again.

5.If you only need to use minute-level scheduled tasks, it is recommended to enable Hhxsv5\LaravelS\Illuminate\LaravelScheduleJob instead of Linux Crontab, so that you can follow the coding habits of Laravel task scheduling and configure Kernel.

// app/Console/Kernel.php
protected function schedule(Schedule $schedule)
{
    // runInBackground() will start a new child process to execute the task. This is asynchronous and will not affect the execution timing of other tasks.
    $schedule->command(TestCommand::class)->runInBackground()->everyMinute();
}

Automatically reload after modifying code

Via inotify, support Linux only.

1.Install inotify extension.

2.Turn on the switch in Settings.

3.Notice: Modify the file only in Linux to receive the file change events. It's recommended to use the latest Docker. Vagrant Solution.

Via fswatch, support OS X/Linux/Windows.

1.Install fswatch.

2.Run command in your project root directory.

# Watch current directory
./bin/fswatch
# Watch app directory
./bin/fswatch ./app

Via inotifywait, support Linux.

1.Install inotify-tools.

2.Run command in your project root directory.

# Watch current directory
./bin/inotify
# Watch app directory
./bin/inotify ./app

When the above methods does not work, the ultimate solution: set max_request=1,worker_num=1, so that Worker process will restart after processing a request. The performance of this method is very poor, so only development environment use.

Get the instance of SwooleServer in your project

/**
 * $swoole is the instance of `Swoole\WebSocket\Server` if enable WebSocket server, otherwise `Swoole\Http\Server`
 * @var \Swoole\WebSocket\Server|\Swoole\Http\Server $swoole
 */
$swoole = app('swoole');
var_dump($swoole->stats());
$swoole->push($fd, 'Push WebSocket message');

Use SwooleTable

1.Define Table, support multiple.

All defined tables will be created before Swoole starting.

// in file "config/laravels.php"
[
    // ...
    'swoole_tables'  => [
        // Scene:bind UserId & FD in WebSocket
        'ws' => [// The Key is table name, will add suffix "Table" to avoid naming conflicts. Here defined a table named "wsTable"
            'size'   => 102400,// The max size
            'column' => [// Define the columns
                ['name' => 'value', 'type' => \Swoole\Table::TYPE_INT, 'size' => 8],
            ],
        ],
        //...Define the other tables
    ],
    // ...
];

2.Access Table: all table instances will be bound on SwooleServer, access by app('swoole')->xxxTable.

namespace App\Services;
use Hhxsv5\LaravelS\Swoole\WebSocketHandlerInterface;
use Swoole\Http\Request;
use Swoole\WebSocket\Frame;
use Swoole\WebSocket\Server;
class WebSocketService implements WebSocketHandlerInterface
{
    /**@var \Swoole\Table $wsTable */
    private $wsTable;
    public function __construct()
    {
        $this->wsTable = app('swoole')->wsTable;
    }
    // Scene:bind UserId & FD in WebSocket
    public function onOpen(Server $server, Request $request)
    {
        // var_dump(app('swoole') === $server);// The same instance
        /**
         * Get the currently logged in user
         * This feature requires that the path to establish a WebSocket connection go through middleware such as Authenticate.
         * E.g:
         * Browser side: var ws = new WebSocket("ws://127.0.0.1:5200/ws");
         * Then the /ws route in Laravel needs to add the middleware like Authenticate.
         * Route::get('/ws', function () {
         *     // Respond any content with status code 200
         *     return 'websocket';
         * })->middleware(['auth']);
         */
        // $user = Auth::user();
        // $userId = $user ? $user->id : 0; // 0 means a guest user who is not logged in
        $userId = mt_rand(1000, 10000);
        // if (!$userId) {
        //     // Disconnect the connections of unlogged users
        //     $server->disconnect($request->fd);
        //     return;
        // }
        $this->wsTable->set('uid:' . $userId, ['value' => $request->fd]);// Bind map uid to fd
        $this->wsTable->set('fd:' . $request->fd, ['value' => $userId]);// Bind map fd to uid
        $server->push($request->fd, "Welcome to LaravelS #{$request->fd}");
    }
    public function onMessage(Server $server, Frame $frame)
    {
        // Broadcast
        foreach ($this->wsTable as $key => $row) {
            if (strpos($key, 'uid:') === 0 && $server->isEstablished($row['value'])) {
                $content = sprintf('Broadcast: new message "%s" from #%d', $frame->data, $frame->fd);
                $server->push($row['value'], $content);
            }
        }
    }
    public function onClose(Server $server, $fd, $reactorId)
    {
        $uid = $this->wsTable->get('fd:' . $fd);
        if ($uid !== false) {
            $this->wsTable->del('uid:' . $uid['value']); // Unbind uid map
        }
        $this->wsTable->del('fd:' . $fd);// Unbind fd map
        $server->push($fd, "Goodbye #{$fd}");
    }
}

Multi-port mixed protocol

For more information, please refer to Swoole Server AddListener

To make our main server support more protocols not just Http and WebSocket, we bring the feature multi-port mixed protocol of Swoole in LaravelS and name it Socket. Now, you can build TCP/UDP applications easily on top of Laravel.

Create Socket handler class, and extend Hhxsv5\LaravelS\Swoole\Socket\{TcpSocket|UdpSocket|Http|WebSocket}.

namespace App\Sockets;
use Hhxsv5\LaravelS\Swoole\Socket\TcpSocket;
use Swoole\Server;
class TestTcpSocket extends TcpSocket
{
    public function onConnect(Server $server, $fd, $reactorId)
    {
        \Log::info('New TCP connection', [$fd]);
        $server->send($fd, 'Welcome to LaravelS.');
    }
    public function onReceive(Server $server, $fd, $reactorId, $data)
    {
        \Log::info('Received data', [$fd, $data]);
        $server->send($fd, 'LaravelS: ' . $data);
        if ($data === "quit\r\n") {
            $server->send($fd, 'LaravelS: bye' . PHP_EOL);
            $server->close($fd);
        }
    }
    public function onClose(Server $server, $fd, $reactorId)
    {
        \Log::info('Close TCP connection', [$fd]);
        $server->send($fd, 'Goodbye');
    }
}

These Socket connections share the same worker processes with your HTTP/WebSocket connections. So it won't be a problem at all if you want to deliver tasks, use SwooleTable, even Laravel components such as DB, Eloquent and so on. At the same time, you can access Swoole\Server\Port object directly by member property swoolePort.

public function onReceive(Server $server, $fd, $reactorId, $data)
{
    $port = $this->swoolePort; // Get the `Swoole\Server\Port` object
}
namespace App\Http\Controllers;
class TestController extends Controller
{
    public function test()
    {
        /**@var \Swoole\Http\Server|\Swoole\WebSocket\Server $swoole */
        $swoole = app('swoole');
        // $swoole->ports: Traverse all Port objects, https://www.swoole.co.uk/docs/modules/swoole-server/multiple-ports
        $port = $swoole->ports[0]; // Get the `Swoole\Server\Port` object, $port[0] is the port of the main server
        foreach ($port->connections as $fd) { // Traverse all connections
            // $swoole->send($fd, 'Send tcp message');
            // if($swoole->isEstablished($fd)) {
            //     $swoole->push($fd, 'Send websocket message');
            // }
        }
    }
}

Register Sockets.

// Edit `config/laravels.php`
//...
'sockets' => [
    [
        'host'     => '127.0.0.1',
        'port'     => 5291,
        'type'     => SWOOLE_SOCK_TCP,// Socket type: SWOOLE_SOCK_TCP/SWOOLE_SOCK_TCP6/SWOOLE_SOCK_UDP/SWOOLE_SOCK_UDP6/SWOOLE_UNIX_DGRAM/SWOOLE_UNIX_STREAM
        'settings' => [// Swoole settings:https://www.swoole.co.uk/docs/modules/swoole-server-methods#swoole_server-addlistener
            'open_eof_check' => true,
            'package_eof'    => "\r\n",
        ],
        'handler'  => \App\Sockets\TestTcpSocket::class,
        'enable'   => true, // whether to enable, default true
    ],
],

About the heartbeat configuration, it can only be set on the main server and cannot be configured on Socket, but the Socket inherits the heartbeat configuration of the main server.

For TCP socket, onConnect and onClose events will be blocked when dispatch_mode of Swoole is 1/3, so if you want to unblock these two events please set dispatch_mode to 2/4/5.

'swoole' => [
    //...
    'dispatch_mode' => 2,
    //...
];

Test.

TCP: telnet 127.0.0.1 5291

UDP: [Linux] echo "Hello LaravelS" > /dev/udp/127.0.0.1/5292

Register example of other protocols.

  • UDP
'sockets' => [
    [
        'host'     => '0.0.0.0',
        'port'     => 5292,
        'type'     => SWOOLE_SOCK_UDP,
        'settings' => [
            'open_eof_check' => true,
            'package_eof'    => "\r\n",
        ],
        'handler'  => \App\Sockets\TestUdpSocket::class,
    ],
],
  • Http
'sockets' => [
    [
        'host'     => '0.0.0.0',
        'port'     => 5293,
        'type'     => SWOOLE_SOCK_TCP,
        'settings' => [
            'open_http_protocol' => true,
        ],
        'handler'  => \App\Sockets\TestHttp::class,
    ],
],
  • WebSocket: The main server must turn on WebSocket, that is, set websocket.enable to true.
'sockets' => [
    [
        'host'     => '0.0.0.0',
        'port'     => 5294,
        'type'     => SWOOLE_SOCK_TCP,
        'settings' => [
            'open_http_protocol'      => true,
            'open_websocket_protocol' => true,
        ],
        'handler'  => \App\Sockets\TestWebSocket::class,
    ],
],

Coroutine

Swoole Coroutine

Warning: The order of code execution in the coroutine is out of order. The data of the request level should be isolated by the coroutine ID. However, there are many singleton and static attributes in Laravel/Lumen, the data between different requests will affect each other, it's Unsafe. For example, the database connection is a singleton, the same database connection shares the same PDO resource. This is fine in the synchronous blocking mode, but it does not work in the asynchronous coroutine mode. Each query needs to create different connections and maintain IO state of different connections, which requires a connection pool.

DO NOT enable the coroutine, only the custom process can use the coroutine.

Custom process

Support developers to create special work processes for monitoring, reporting, or other special tasks. Refer addProcess.

Create Proccess class, implements CustomProcessInterface.

namespace App\Processes;
use App\Tasks\TestTask;
use Hhxsv5\LaravelS\Swoole\Process\CustomProcessInterface;
use Hhxsv5\LaravelS\Swoole\Task\Task;
use Swoole\Coroutine;
use Swoole\Http\Server;
use Swoole\Process;
class TestProcess implements CustomProcessInterface
{
    /**
     * @var bool Quit tag for Reload updates
     */
    private static $quit = false;

    public static function callback(Server $swoole, Process $process)
    {
        // The callback method cannot exit. Once exited, Manager process will automatically create the process 
        while (!self::$quit) {
            \Log::info('Test process: running');
            // sleep(1); // Swoole < 2.1
            Coroutine::sleep(1); // Swoole>=2.1: Coroutine & Runtime will be automatically enabled for callback().
             // Deliver task in custom process, but NOT support callback finish() of task.
            // Note: Modify task_ipc_mode to 1 or 2 in config/laravels.php, see https://www.swoole.co.uk/docs/modules/swoole-server/configuration
            $ret = Task::deliver(new TestTask('task data'));
            var_dump($ret);
            // The upper layer will catch the exception thrown in the callback and record it in the Swoole log, and then this process will exit. The Manager process will re-create the process after 3 seconds, so developers need to try/catch to catch the exception by themselves to avoid frequent process creation.
            // throw new \Exception('an exception');
        }
    }
    // Requirements: LaravelS >= v3.4.0 & callback() must be async non-blocking program.
    public static function onReload(Server $swoole, Process $process)
    {
        // Stop the process...
        // Then end process
        \Log::info('Test process: reloading');
        self::$quit = true;
        // $process->exit(0); // Force exit process
    }
    // Requirements: LaravelS >= v3.7.4 & callback() must be async non-blocking program.
    public static function onStop(Server $swoole, Process $process)
    {
        // Stop the process...
        // Then end process
        \Log::info('Test process: stopping');
        self::$quit = true;
        // $process->exit(0); // Force exit process
    }
}

Register TestProcess.

// Edit `config/laravels.php`
// ...
'processes' => [
    'test' => [ // Key name is process name
        'class'    => \App\Processes\TestProcess::class,
        'redirect' => false, // Whether redirect stdin/stdout, true or false
        'pipe'     => 0,     // The type of pipeline, 0: no pipeline 1: SOCK_STREAM 2: SOCK_DGRAM
        'enable'   => true,  // Whether to enable, default true
        //'num'    => 3   // To create multiple processes of this class, default is 1
        //'queue'    => [ // Enable message queue as inter-process communication, configure empty array means use default parameters
        //    'msg_key'  => 0,    // The key of the message queue. Default: ftok(__FILE__, 1).
        //    'mode'     => 2,    // Communication mode, default is 2, which means contention mode
        //    'capacity' => 8192, // The length of a single message, is limited by the operating system kernel parameters. The default is 8192, and the maximum is 65536
        //],
        //'restart_interval' => 5, // After the process exits abnormally, how many seconds to wait before restarting the process, default 5 seconds
    ],
],

Note: The callback() cannot quit. If quit, the Manager process will re-create the process.

Example: Write data to a custom process.

// config/laravels.php
'processes' => [
    'test' => [
        'class'    => \App\Processes\TestProcess::class,
        'redirect' => false,
        'pipe'     => 1,
    ],
],
// app/Processes/TestProcess.php
public static function callback(Server $swoole, Process $process)
{
    while ($data = $process->read()) {
        \Log::info('TestProcess: read data', [$data]);
        $process->write('TestProcess: ' . $data);
    }
}
// app/Http/Controllers/TestController.php
public function testProcessWrite()
{
    /**@var \Swoole\Process $process */
    $process = app('swoole')->customProcesses['test'];
    $process->write('TestController: write data' . time());
    var_dump($process->read());
}

Common components

Apollo

LaravelS will pull the Apollo configuration and write it to the .env file when starting. At the same time, LaravelS will start the custom process apollo to monitor the configuration and automatically reload when the configuration changes.

Enable Apollo: add --enable-apollo and Apollo parameters to the startup parameters.

php bin/laravels start --enable-apollo --apollo-server=http://127.0.0.1:8080 --apollo-app-id=LARAVEL-S-TEST

Support hot updates(optional).

// Edit `config/laravels.php`
'processes' => Hhxsv5\LaravelS\Components\Apollo\Process::getDefinition(),
// When there are other custom process configurations
'processes' => [
    'test' => [
        'class'    => \App\Processes\TestProcess::class,
        'redirect' => false,
        'pipe'     => 1,
    ],
    // ...
] + Hhxsv5\LaravelS\Components\Apollo\Process::getDefinition(),

List of available parameters.

ParameterDescriptionDefaultDemo
apollo-serverApollo server URL---apollo-server=http://127.0.0.1:8080
apollo-app-idApollo APP ID---apollo-app-id=LARAVEL-S-TEST
apollo-namespacesThe namespace to which the APP belongs, support specify the multipleapplication--apollo-namespaces=application --apollo-namespaces=env
apollo-clusterThe cluster to which the APP belongsdefault--apollo-cluster=default
apollo-client-ipIP of current instance, can also be used for grayscale publishingLocal intranet IP--apollo-client-ip=10.2.1.83
apollo-pull-timeoutTimeout time(seconds) when pulling configuration5--apollo-pull-timeout=5
apollo-backup-old-envWhether to backup the old configuration file when updating the configuration file .envfalse--apollo-backup-old-env

Prometheus

Support Prometheus monitoring and alarm, Grafana visually view monitoring metrics. Please refer to Docker Compose for the environment construction of Prometheus and Grafana.

Require extension APCu >= 5.0.0, please install it by pecl install apcu.

Copy the configuration file prometheus.php to the config directory of your project. Modify the configuration as appropriate.

# Execute commands in the project root directory
cp vendor/hhxsv5/laravel-s/config/prometheus.php config/

If your project is Lumen, you also need to manually load the configuration $app->configure('prometheus'); in bootstrap/app.php.

Configure global middleware: Hhxsv5\LaravelS\Components\Prometheus\RequestMiddleware::class. In order to count the request time consumption as accurately as possible, RequestMiddleware must be the first global middleware, which needs to be placed in front of other middleware.

Register ServiceProvider: Hhxsv5\LaravelS\Components\Prometheus\ServiceProvider::class.

Configure the CollectorProcess in config/laravels.php to collect the metrics of Swoole Worker/Task/Timer processes regularly.

'processes' => Hhxsv5\LaravelS\Components\Prometheus\CollectorProcess::getDefinition(),

Create the route to output metrics.

use Hhxsv5\LaravelS\Components\Prometheus\Exporter;

Route::get('/actuator/prometheus', function () {
    $result = app(Exporter::class)->render();
    return response($result, 200, ['Content-Type' => Exporter::REDNER_MIME_TYPE]);
});

Complete the configuration of Prometheus and start it.

global:
  scrape_interval: 5s
  scrape_timeout: 5s
  evaluation_interval: 30s
scrape_configs:
- job_name: laravel-s-test
  honor_timestamps: true
  metrics_path: /actuator/prometheus
  scheme: http
  follow_redirects: true
  static_configs:
  - targets:
    - 127.0.0.1:5200 # The ip and port of the monitored service
# Dynamically discovered using one of the supported service-discovery mechanisms
# https://prometheus.io/docs/prometheus/latest/configuration/configuration/#scrape_config
# - job_name: laravels-eureka
#   honor_timestamps: true
#   scrape_interval: 5s
#   metrics_path: /actuator/prometheus
#   scheme: http
#   follow_redirects: true
  # eureka_sd_configs:
  # - server: http://127.0.0.1:8080/eureka
  #   follow_redirects: true
  #   refresh_interval: 5s

Start Grafana, then import panel json.

Grafana Dashboard

Other features

Configure Swoole events

Supported events:

EventInterfaceWhen happened
ServerStartHhxsv5\LaravelS\Swoole\Events\ServerStartInterfaceOccurs when the Master process is starting, this event should not handle complex business logic, and can only do some simple work of initialization.
ServerStopHhxsv5\LaravelS\Swoole\Events\ServerStopInterfaceOccurs when the server exits normally, CANNOT use async or coroutine related APIs in this event.
WorkerStartHhxsv5\LaravelS\Swoole\Events\WorkerStartInterfaceOccurs after the Worker/Task process is started, and the Laravel initialization has been completed.
WorkerStopHhxsv5\LaravelS\Swoole\Events\WorkerStopInterfaceOccurs after the Worker/Task process exits normally
WorkerErrorHhxsv5\LaravelS\Swoole\Events\WorkerErrorInterfaceOccurs when an exception or fatal error occurs in the Worker/Task process

1.Create an event class to implement the corresponding interface.

namespace App\Events;
use Hhxsv5\LaravelS\Swoole\Events\ServerStartInterface;
use Swoole\Atomic;
use Swoole\Http\Server;
class ServerStartEvent implements ServerStartInterface
{
    public function __construct()
    {
    }
    public function handle(Server $server)
    {
        // Initialize a global counter (available across processes)
        $server->atomicCount = new Atomic(2233);

        // Invoked in controller: app('swoole')->atomicCount->get();
    }
}
namespace App\Events;
use Hhxsv5\LaravelS\Swoole\Events\WorkerStartInterface;
use Swoole\Http\Server;
class WorkerStartEvent implements WorkerStartInterface
{
    public function __construct()
    {
    }
    public function handle(Server $server, $workerId)
    {
        // Initialize a database connection pool
        // DatabaseConnectionPool::init();
    }
}

2.Configuration.

// Edit `config/laravels.php`
'event_handlers' => [
    'ServerStart' => [\App\Events\ServerStartEvent::class], // Trigger events in array order
    'WorkerStart' => [\App\Events\WorkerStartEvent::class],
],

Serverless

Alibaba Cloud Function Compute

Function Compute.

1.Modify bootstrap/app.php and set the storage directory. Because the project directory is read-only, the /tmp directory can only be read and written.

$app->useStoragePath(env('APP_STORAGE_PATH', '/tmp/storage'));

2.Create a shell script laravels_bootstrap and grant executable permission.

#!/usr/bin/env bash
set +e

# Create storage-related directories
mkdir -p /tmp/storage/app/public
mkdir -p /tmp/storage/framework/cache
mkdir -p /tmp/storage/framework/sessions
mkdir -p /tmp/storage/framework/testing
mkdir -p /tmp/storage/framework/views
mkdir -p /tmp/storage/logs

# Set the environment variable APP_STORAGE_PATH, please make sure it's the same as APP_STORAGE_PATH in .env
export APP_STORAGE_PATH=/tmp/storage

# Start LaravelS
php bin/laravels start

3.Configure template.xml.

ROSTemplateFormatVersion: '2015-09-01'
Transform: 'Aliyun::Serverless-2018-04-03'
Resources:
  laravel-s-demo:
    Type: 'Aliyun::Serverless::Service'
    Properties:
      Description: 'LaravelS Demo for Serverless'
    fc-laravel-s:
      Type: 'Aliyun::Serverless::Function'
      Properties:
        Handler: laravels.handler
        Runtime: custom
        MemorySize: 512
        Timeout: 30
        CodeUri: ./
        InstanceConcurrency: 10
        EnvironmentVariables:
          BOOTSTRAP_FILE: laravels_bootstrap

Important notices

Singleton Issue

Under FPM mode, singleton instances will be instantiated and recycled in every request, request start=>instantiate instance=>request end=>recycled instance.

Under Swoole Server, All singleton instances will be held in memory, different lifetime from FPM, request start=>instantiate instance=>request end=>do not recycle singleton instance. So need developer to maintain status of singleton instances in every request.

Common solutions:

Write a XxxCleaner class to clean up the singleton object state. This class implements the interface Hhxsv5\LaravelS\Illuminate\Cleaners\CleanerInterface and then registers it in cleaners of laravels.php.

Reset status of singleton instances by Middleware.

Re-register ServiceProvider, add XxxServiceProvider into register_providers of file laravels.php. So that reinitialize singleton instances in every request Refer.

Cleaners

Configuration cleaners.

Known issues

Known issues: a package of known issues and solutions.

Debugging method

Logging; if you want to output to the console, you can use stderr, Log::channel('stderr')->debug('debug message').

Laravel Dump Server(Laravel 5.7 has been integrated by default).

Read request

Read request by Illuminate\Http\Request Object, $_ENV is readable, $_SERVER is partially readable, CANNOT USE $_GET/$_POST/$_FILES/$_COOKIE/$_REQUEST/$_SESSION/$GLOBALS.

public function form(\Illuminate\Http\Request $request)
{
    $name = $request->input('name');
    $all = $request->all();
    $sessionId = $request->cookie('sessionId');
    $photo = $request->file('photo');
    // Call getContent() to get the raw POST body, instead of file_get_contents('php://input')
    $rawContent = $request->getContent();
    //...
}

Output response

Respond by Illuminate\Http\Response Object, compatible with echo/vardump()/print_r(),CANNOT USE functions dd()/exit()/die()/header()/setcookie()/http_response_code().

public function json()
{
    return response()->json(['time' => time()])->header('header1', 'value1')->withCookie('c1', 'v1');
}

Persistent connection

Singleton connection will be resident in memory, it is recommended to turn on persistent connection for better performance.

  1. Database connection, it will reconnect automatically immediately after disconnect.
// config/database.php
'connections' => [
    'my_conn' => [
        'driver'    => 'mysql',
        'host'      => env('DB_MY_CONN_HOST', 'localhost'),
        'port'      => env('DB_MY_CONN_PORT', 3306),
        'database'  => env('DB_MY_CONN_DATABASE', 'forge'),
        'username'  => env('DB_MY_CONN_USERNAME', 'forge'),
        'password'  => env('DB_MY_CONN_PASSWORD', ''),
        'charset'   => 'utf8mb4',
        'collation' => 'utf8mb4_unicode_ci',
        'prefix'    => '',
        'strict'    => false,
        'options'   => [
            // Enable persistent connection
            \PDO::ATTR_PERSISTENT => true,
        ],
    ],
],
  1. Redis connection, it won't reconnect automatically immediately after disconnect, and will throw an exception about lost connection, reconnect next time. You need to make sure that SELECT DB correctly before operating Redis every time.
// config/database.php
'redis' => [
    'client' => env('REDIS_CLIENT', 'phpredis'), // It is recommended to use phpredis for better performance.
    'default' => [
        'host'       => env('REDIS_HOST', 'localhost'),
        'password'   => env('REDIS_PASSWORD', null),
        'port'       => env('REDIS_PORT', 6379),
        'database'   => 0,
        'persistent' => true, // Enable persistent connection
    ],
],

About memory leaks

Avoid using global variables. If necessary, please clean or reset them manually.

Infinitely appending element into static/global variable will lead to OOM(Out of Memory).

class Test
{
    public static $array = [];
    public static $string = '';
}

// Controller
public function test(Request $req)
{
    // Out of Memory
    Test::$array[] = $req->input('param1');
    Test::$string .= $req->input('param2');
}

Memory leak detection method

Modify config/laravels.php: worker_num=1, max_request=1000000, remember to change it back after test;

Add routing /debug-memory-leak without route middleware to observe the memory changes of the Worker process;

Route::get('/debug-memory-leak', function () {
    global $previous;
    $current = memory_get_usage();
    $stats = [
        'prev_mem' => $previous,
        'curr_mem' => $current,
        'diff_mem' => $current - $previous,
    ];
    $previous = $current;
    return $stats;
});

Start LaravelS and request /debug-memory-leak until diff_mem is less than or equal to zero; if diff_mem is always greater than zero, it means that there may be a memory leak in Global Middleware or Laravel Framework;

After completing Step 3, alternately request the business routes and /debug-memory-leak (It is recommended to use ab/wrk to make a large number of requests for business routes), the initial increase in memory is normal. After a large number of requests for the business routes, if diff_mem is always greater than zero and curr_mem continues to increase, there is a high probability of memory leak; If curr_mem always changes within a certain range and does not continue to increase, there is a low probability of memory leak.

If you still can't solve it, max_request is the last guarantee.

Linux kernel parameter adjustment

Linux kernel parameter adjustment

Pressure test

Pressure test


Author: hhxsv5
Source Code: https://github.com/hhxsv5/laravel-s
License: MIT License

#laravel #php