Welcome to Advent of Code 2024!

Like every year, I start the challenge with the best attitude and love of being an Elixir programmer. Although I know that at some point, I will go to the “what is this? I hate it” phase, unlike other years, this time, I am committed to finishing Advent of Code and, more importantly, sharing it with you.

I hope you enjoy this series of December posts, where we will discuss the approach for each exercise. But remember that it is not the only one, and the idea of ​​this initiative is to have a great time and share knowledge, so don’t forget to post your solutions and comments and tag us to continue the conversation.

Let’s go for it!

Day 1: Historian Hysteria

Before starting any exercise, I suggest spending some time defining the structure that best fits the problem’s needs. If the structure is adequate, it will be easy to reuse it for the second part without further complications.

In this case, the exercise itself describes lists as the input, so we can skip that step and instead consider which functions of the Enum or List modules can be helpful.

We have this example input:

3 4

4 3

2 5

1 3

3 9

3   3

The goal is to transform it into two separate lists and apply sorting, comparison, etc.

List 1: [3, 4, 2, 1, 3, 3 ]

List 2: [ 4, 3, 5, 3, 9, 3 ]

Let’s define a function that reads a file with the input. Each line will initially be represented by a string, so use String . split to separate it at each line break.

 def get_input(path) do
   path
   |> File.read!()
   |> String.split("\n", trim: true)
 end


["3   4", "4   3", "2   5", "1   3", "3   9", "3   3"]

We will still have each row represented by a string, but we can now modify this using the functions in the Enum module. Notice that the whitespace between characters is constant, and the pattern is that the first element should go into list one and the second element into list two. Use Enum.reduce to map the elements to the corresponding list and get the following output:

%{
 first_list: [3, 3, 1, 2, 4, 3],
 second_list: [3, 9, 3, 5, 3, 4]
}

I’m using a map so that we can identify the lists and everything is clear. The function that creates them is as follows:

 @doc """
 This function takes a list where the elements are strings with two
 components separated by whitespace.


 Example: "3   4"


 It assigns the first element to list one and the second to list two,
 assuming both are numbers.
 """
 def define_separated_lists(input) do
   Enum.reduce(input, %{first_list: [], second_list: []}, fn row, map_with_lists ->
     [elem_first_list, elem_second_list] = String.split(row, "   ")


     %{
       first_list: [String.to_integer(elem_first_list) | map_with_lists.first_list],
       second_list: [String.to_integer(elem_second_list) | map_with_lists.second_list]
     }
   end)
 end

Once we have this format, we can move on to the first part of the exercise.

Part 1

Use Enum.sort to sort the lists ascendingly and pass them to the Enum.zip_with function that will calculate the distance between the elements of both. Note that we are using abs to avoid negative values, and finally, Enum.reduce to sum all the distances.

first_sorted_list = Enum.sort(first_list)
   second_sorted_list = Enum.sort(second_list)


   first_sorted_list
   |> Enum.zip_with(second_sorted_list, fn x, y -> abs(x-y) end)
   |> Enum.reduce(0, fn distance, acc -> distance + acc end)

Part 2

For the second part, you don’t need to sort the lists; use Enum. frequencies and Enum.reduce to get the multiplication of the elements.

 frequencies_second_list = Enum.frequencies(second_list)


   Enum.reduce(first_list, 0, fn elem, acc ->
     elem * Map.get(frequencies_second_list, elem, 0) + acc
   end)

That’s it. As you can see, once we have a good structure, the corresponding module, in this case, Enum, makes the operations more straightforward, so it’s worth spending some time defining which input will make our life easier.

You can see the full version of the exercise here .

The post Advent of Code 2024 appeared first on Erlang Solutions .

The success of any programming language in the Erlang ecosystem can be apportioned into three tightly coupled components. They are the semantics of the Erlang programming language , (on top of which other languages are implemented), the OTP libraries and middleware (used to architect scalable and resilient concurrent systems) and the BEAM Virtual Machine tightly coupled to the language semantics and OTP.

Take any of these components on their own, and you have a runner-up. But, put the three together, and you have the uncontested winner for scalable, resilient soft-real real-time systems. To quote Joe Armstrong, “You can copy the Erlang libraries, but if it does not run on BEAM, you can’t emulate the semantics”. This is enforced by Robert Virding’s First Rule of Programming, which states that “Any sufficiently complicated concurrent program in another language contains an ad hoc informally-specified bug-ridden slow implementation of half of Erlang.”

In this post, we want to explore the BEAM VM internals. We will compare and contrast them with the JVM where applicable, highlighting why you should pay attention to them and care. For too long, this component has been treated as a black box and taken for granted, without understanding the reasons or implications. It is time to change that!

Highlights of the BEAM

Erlang and the BEAM VM were invented to be the right tools to solve a specific problem. Ericsson developed them to help implement telecom infrastructure, handling both mobile and fixed networks. This infrastructure is highly concurrent and scalable in nature. It has to display soft real-time properties and may never fail. We don’t want our phone calls dropped or our online gaming experience to be affected by system upgrades, high user load or software, hardware and network outages. The BEAM VM solves these challenges using a state-of-the-art concurrent programming model. It features lightweight BEAM processes which don’t share memory, are managed by the schedulers of the BEAM which can manage millions of them across multiple cores, and garbage collectors running on a per-process basis, highly optimised to reduce any impact on other processes. The BEAM is also the only widely used VM used at scale with a built-in distribution model which allows a program to run on multiple machines transparently.

The BEAM VM supports zero-downtime upgrades with hot code replacement , a way to modify application code at runtime. It is probably the most cited unique feature of the BEAM. Hot code loading means that the application logic can be updated by changing the runnable code in the system whilst retaining the internal process state. This is achieved by replacing the loaded BEAM files and instructing the VM to replace the references of the code in the running processes.

It is a crucial feature for no downtime code upgrades for telecom infrastructure, where redundant hardware was put to use to handle spikes. Nowadays, in the era of containerisation, other techniques are also used for production updates. Those who have never used it dismiss it as a less important feature, but it is nonetheless useful in the development workflow. Developers can iterate faster by replacing part of their code without having to restart the system to test it. Even if the application is not designed to be upgradable in production, this can reduce the time needed for recompilation and redeployments.

Highlights of the JVM

The Java Virtual Machine (JVM) was invented by Sun Microsystem with the intent to provide a platform for ‘write once’ code that runs everywhere. They created an object-oriented language similar to C++ , but memory-safe because its runtime error detection checks array bounds and pointer dereferences. The JVM ecosystem became extremely popular in the Internet era, making it the de-facto standard for enterprise server applications. The wide range of applicability was enabled by a virtual machine that caters for many use cases, and an impressive set of libraries supporting enterprise development.

The JVM was designed with efficiency in mind. Most of its concepts are abstractions of features found in popular operating systems such as the threading model which maps the VM threads to operating system threads. The JVM is highly customisable, including the garbage collector (GC) and class loaders. Some state-of-the-art GC implementations provide highly tunable features catering for a programming model based on shared memory. And, the JIT (Just-in-time) compiler automatically compiles bytecode to native machine code with the intent to speed up parts of the application.

The JVM allows you to change the code while the program is running. It is a very useful feature for debugging purposes, but production use of this feature is not recommended due to serious limitations .

Concurrency and Parallelism

We talk about parallel code execution if parts of the code are run at the same time on multiple cores, processors or computers, while concurrent programming refers to handling events arriving at the system independently. Concurrent execution can be simulated on single-core hardware, while parallel execution cannot. Although this distinction may seem pedantic, the difference results in some very different problems to solve. Think of many cooks making a plate of carbonara pasta. In the parallel approach, the tasks are split across the number of cooks available, and a single portion would be completed as quickly as it took these cooks to complete their specific tasks. In a concurrent world, you would get a portion for every cook, where each cook does all of the tasks. You use parallelism for speed and concurrency for scale.

Parallel execution tries to decompose the problem into parts that are independent of each other. Boil the water, get the pasta, mix the egg, fry the guanciale ham, and grate the pecorino cheese ¹ . The shared data (or in our example, the serving dish) is handled by locks, mutexes and various other techniques to guarantee correctness. Another way to look at this is that the data (or ingredients) are present, and we want to utilise as many parallel CPU resources as possible to finish the job as quickly as possible.

Concurrent programming, on the other hand, deals with many events that arrive at the system at different times and tries to process all of them within a reasonable timeframe. On multi-core or distributed architectures, some of the processing may run in parallel. Another way to look at it is that the same cook boils the water, gets the pasta, mixes the eggs and so on, following a sequential algorithm which is always the same. What changes across processes (or cooks) is the data (or ingredients) to work on, which exist in multiple instances.

In summary, concurrency and parallelism are two intrinsically different problems, requiring different solutions.

Concurrency the Java way

In Java, concurrent execution is implemented using VM threads. Before the latest developments, only one threading model, called Platform Threads existed. As it is a thin abstraction layer above operating system threads, Platform Threads are scheduled in a rather simple, priority-based way, leaving most of the work to the underlying operating system. With Java 21, a new threading model was introduced, the Virtual Threads. This new model is very similar to BEAM processes since virtual threads are scheduled by the JVM, providing better performance in applications where thread contention is not negligible. Scheduling works by mounting a virtual thread to the carrier (OS) thread and unmounting it when the state of the virtual thread becomes blocked, and replacing it with a new virtual thread from the pool.

Since Java promotes the use of shared data structures, both threading models suffer from performance bottlenecks caused by synchronisation-related issues like frequent CPU cache invalidation and locking errors. Also, programming with concurrency primitives is a difficult task because of the challenges created by the shared memory model. To overcome these difficulties, there are attempts to simplify and unify the concurrent programming models, with the most successful attempt being the Akka framework . Unfortunately, it is not widely used, limiting its usefulness as a unified concurrency model, even for enterprise-grade applications. While Akka does a great job at replicating the higher-level constructs, it is somewhat limited by the lack of primitives provided by the JVM, allowing it to be highly optimised for concurrency. While the primitives of the JVM enable a wider range of use cases, they make programming distributed systems harder as they have no built-in primitives for communication and are often based on a shared memory model. For example, wherein a distributed system do you place your shared memory? And what is the cost of accessing it?

Garbage Collection

Garbage collection is a critical task for most of the applications, but applications may have very different performance requirements. Since the JVM is designed to be a ubiquitous platform, it is evident that there is no one-size-fits-all solution. There are garbage collectors designed for resource-limited environments such as embedded devices, and also for resource-intensive, highly concurrent or even real-time applications. The JVM GC interface makes it possible to use 3rd party collectors as well.

Due to the Java Memory Model , concurrent garbage collection is a hard task. The JVM needs to keep track of the memory areas that are shared between multiple threads, the access patterns to the shared memory, thread states, locks and so on. Because of shared memory, collections affect multiple threads simultaneously, making it difficult to predict the performance impact of GC operations. So difficult, that there is an entire industry built to provide tools and expertise for GC optimisation.

The BEAM and Concurrency

Some say that the JVM is built for parallelism, the BEAM for concurrency. While this might be an oversimplification, its concurrency model makes the BEAM more performant in cases where thousands or even millions of concurrent tasks should be processed in a reasonable timeframe.

The BEAM provides lightweight processes to give context to the running code. BEAM processes are different from operating system processes, but they share many concepts. BEAM processes, also called actors, don’t share memory, but communicate through message passing, copying data from one process to another. Message passing is a feature that the virtual machine implements through mailboxes owned by individual processes. It is a non-blocking operation, which means that sending a message to another process is almost instantaneous and the execution of the sender is not blocked during the operation. The messages sent are in the form of immutable data, copied from the stack of the sending process to the mailbox of the receiving one. There are no shared data structures, so this can be achieved without the need for locks and mutexes among the communicating processes, only a lock on the mailbox in case multiple processes send a message to the same recipient in parallel.

Immutable data and message passing together enable the programmer to write processes which work independently of each other and focus on functionality instead of the low-level management of the memory and scheduling of tasks. It turns out that this simple design is effective on both single thread and multiple threads on a local machine running in the same VM and, using the inter-VM communication facilities of the BEAM, across the network and machines running the BEAM. Because the messages are immutable between processes, they can be scheduled to run on another OS thread (or machine) without locking, providing almost linear scaling on distributed, multi-core architectures. The processes are handled in the same way on a local VM as in a cluster of VMs, message sending works transparently regardless of the location of the receiving process.

Processes do not share memory, allowing data replication for resilience and distribution for scale. Having two instances of the same process on a single or more separate machine, state updates can be shared with each other. If one of the processes or machines fails, the other has an up-to-date copy of the data and can continue handling requests without interruption, making the system fault-tolerant. If more than one machine is operational, all the processes can handle requests, giving you scalability. The BEAM provides highly optimised primitives for all of this to work seamlessly, while OTP (the “standard library”) provides the higher level constructs to make the life of the programmers easy.

Scheduler

We mentioned that one of the strongest features of the BEAM is the ability to run concurrent tasks in lightweight processes. Managing these processes is the task of the scheduler.

The scheduler starts, by default, an OS thread for every core and optimises the workload between them. Each process consists of code to be executed and a state which changes over time. The scheduler picks the first process in the run queue that is ready to run, and gives it a certain amount of reductions to execute, where each reduction is the rough equivalent of a BEAM command. Once the process has either run out of reductions, is blocked by I/O, is waiting for a message, or is completed executing its code, the scheduler picks the next process from the run queue and dispatches it. This scheduling technique is called pre-emptive.

We have mentioned the Akka framework many times. Its biggest drawback is the need to annotate the code with scheduling points, as the scheduling is not done at the JVM level. By removing the control from the hands of the programmer, soft real-time properties are preserved and guaranteed, as there is no risk of them accidentally causing process starvation.

The processes can be spread around the available scheduler threads to maximise CPU utilisation. There are many ways to tweak the scheduler but it is rarely needed, only for edge cases, as the default configuration covers most usage patterns.

There is a sensitive topic that frequently pops up regarding schedulers: how to handle Natively Implemented Functions (NIFs) . A NIF is a code snippet written in C, compiled as a library and run in the same memory space as the BEAM for speed. The problem with NIFs is that they are not pre-emptive, and can affect the schedulers. In recent BEAM versions, a new feature, dirty schedulers, was added to give better control for NIFs. Dirty schedulers are separate schedulers that run in different threads to minimise the interruption a NIF can cause in a system. The word dirty refers to the nature of the code that is run by these schedulers.

Garbage Collector

Modern, high-level programming languages today mostly use a garbage collector for memory management. The BEAM languages are no exception. Trusting the virtual machine to handle the resources and manage the memory is very handy when you want to write high-level concurrent code, as it simplifies the task. The underlying implementation of the garbage collector is fairly straightforward and efficient, thanks to the memory model based on an immutable state. Data is copied, not mutated and the fact that processes do not share memory removes any process inter-dependencies, which, as a result, do not need to be managed.

Another feature of the BEAM is that garbage collection is run only when needed, on a per-process basis, without affecting other processes waiting in the run queue. As a result, the garbage collection in Erlang does not ‘stop the world’. It prevents processing latency spikes because the VM is never stopped as a whole – only specific processes are, and never all of them at the same time. In practice, it is just part of what a process does and is treated as another reduction. The garbage collector collecting process suspends the process for a very short interval, often microseconds. As a result, there will be many small bursts, triggered only when the process needs more memory. A single process usually doesn’t allocate large amounts of memory, and is often short-lived, further reducing the impact by immediately freeing up all its allocated memory on termination.

More about features

The features of the garbage collector are discussed in an excellent blog post by Lukas Larsson . There are many intricate details, but it is optimised to handle immutable data in an efficient way, dividing the data between the stack and the heap for each process. The best approach is to do the majority of the work in short-lived processes.

A question that often comes up on this topic is how much memory the BEAM uses. Under the hood, the VM allocates big chunks of memory and uses custom allocators to store the data efficiently and minimise the overhead of system calls.

This has two visible effects: The used memory decreases gradually after the space is not needed, and reallocating huge amounts of data might mean doubling the current working memory. The first effect can, if necessary, be mitigated by tweaking the allocator strategies. The second one is easy to monitor and plan for if you have visibility of the different types of memory usage. (One such monitoring tool that provides system metrics that are out of the box is WombatOAM ).

JVM vs BEAM concurrency

As mentioned before, the JVM and the BEAM handle concurrent tasks very differently. Under high load, shared resources become bottlenecks. In a Java application, we usually can’t avoid that. That’s why the BEAM is superior in these kinds of applications. While memory copy has a certain cost, the performance impact caused by the synchronised access to shared resources is much higher. We performed many tests to measure this impact.

This chart nicely displays the large differences in performance between the JVM and the BEAM. In this test, the applications were implemented in Elixir and Java. The Elixir code compiles to the BEAM, while the Java code, evidently, compiles to the JVM.

When not to use the BEAM

It is very much about the right tool for the job. Do you need a system to be extremely fast, but are not concerned about concurrency? Handling a few events in parallel, and having to handle them fast? Need to crunch numbers for graphics, AI or analytics? Go down the C++, Python or Java route. Telecom infrastructure does not need fast operations on floats, so speed was never a priority. Aided with dynamic typing, which has to do all type checks at runtime means compile-time optimizations are not as trivial. So number crunching is best left to the JVM, Go or other languages which compile to native code. It is no surprise that floating point operations on Erjang, the version of Erlang running on the JVM, was 5000% faster than the BEAM. But where we’ve seen the BEAM shine is using its concurrency to orchestrate number crunching, outsourcing the analytics to C, Julia, Python or Rust. You do the map outside the BEAM and the reduction within the BEAM.

The mantra has always been fast enough. It takes a few hundred milliseconds for humans to perceive a stimulus (an event) and process it in their brains, meaning that micro or nano-second response time is not necessary for many applications. Nor would you use the BEAM for microcontrollers, it is too resource-hungry. But for embedded systems with a bit more processing power, where multi-core is becoming the norm and you need concurrency, the BEAM shines. Back in the 90s, we were implementing telephony switches handling tens of thousands of subscribers running in embedded boards with 16 MB of memory. How much memory does a Raspberry Pi have these days? And finally, hard real-time systems. You would probably not want the BEAM to manage your airbag control system. You need hard guarantees, something only a hard real-time OS and a language with no garbage collection or exceptions. An implementation of an Erlang VM running on bare metal such as GRiSP will give you similar guarantees.

Conclusion

Use the right tool for the job. If you are writing a soft real-time system which has to scale out of the box, should never fail, and do so without the hassle of having to reinvent the wheel, the BEAM is the battle-proven technology you are looking for.

For many, it works as a black box. Not knowing how it works would be analogous to driving a Ferrari and not being capable of achieving optimal performance or not understanding what part of the motor that strange sound is coming from. This is why you should learn more about the BEAM, understand its internals and be ready to fine-tune and fix it.

For those who have used Erlang and Elixir in anger, we have launched a one-day instructor-led course which will demystify and explain a lot of what you saw whilst preparing you to handle massive concurrency at scale. The course is available through our new instructor-led lead remote training, learn more here . We also recommend The BEAM book by Erik Stenman and the BEAM Wisdoms , a collection of articles by Dmytro Lytovchenko.

If you’d like to speak to a member of the team, feel free to drop us a message .

The post Optimising for Concurrency: Comparing and contrasting the BEAM and JVM virtual machines appeared first on Erlang Solutions .

In an annual vote, not one, not two, but three Ignite Realtime community members have been selected into leadership positions of the XMPP Standards Foundation!

The XMPP Standards Foundation is an independent, nonprofit standards development organisation whose primary mission is to define open protocols for presence, instant messaging, and real-time communication and collaboration on top of the IETF’s Extensible Messaging and Presence Protocol (XMPP). Most of the projects that we’re maintaining in the Ignite Realtime community have a strong dependency on XMPP.

The XSF Board of Directors, in which both @Flow and @dwd are elected, oversees the business affairs of the organisation. They are now in a position to make key decisions on the direction of XMPP technology and standards development, manage resources and partnerships to further the growth of the XMPP ecosystem and promote XMPP in the larger open-source and communications community, advocating for its adoption and use in various applications.

The XMPP Council, in which @danc has been reelected, is the technical steering group that approves XMPP Extension Protocols. The Council is responsible for standards development and process management. With that, Dan is now on the forefront of new developments within the XMPP community!

Congrats to you all, Dan, Dave and Florian!

For other release announcements and news follow us on Mastodon or X

1 post - 1 participant

Read full topic

MongooseIM is a scalable, efficient, high-performance instant messaging server using the proven, open, and extensible XMPP protocol. With each new version, we introduce new features and improvements. For example, version 6.2.0 introduced our new CETS in-memory storage, making setup and autoscaling in cloud environments easier than before (see the blog post for details). The latest release 6.3.0 is no exception. The main highlight is the complete instrumentation rework, allowing seamless integration with modern monitoring solutions like Prometheus.

Additionally, we have added CockroachDB to the list of supported databases, so you can now let this highly scalable database grow with your applications while avoiding being locked into your cloud provider.

Observability and instrumentation

In software engineering, observability is the ability to gather data from a running system to figure out what is going inside: is it working as expected? Does it have any issues? How much load is it handling, and could it do more? There are many ways to improve the observability of a system, and one of the most important is instrumentation . Just like adding extra measuring equipment to a physical system, this means adding additional code to the software. It allows the system administrator to observe the internal state of the system. This comes with a price. There is more work for the developers, increased complexity, and potential performance degradation caused by the collection and processing of additional data.

However, the benefits usually outweigh the costs, and the ability to inspect the system is often a critical requirement. It is also worth noting that the metrics and events gathered by instrumentation can be used for further automation, e.g. for autoscaling or sending alarms to the administrator.

Instrumentation in MongooseIM

Even before our latest release of MongooseIM, there have been multiple means to observe its behaviour:

Metrics provide numerical values of measured system properties. The values change over time, and the metric can present current value, sum from a sliding window, or a statistic (histogram) of values from a given time period. Prior to version 6.3, MongooseIM used to store such metrics with the help of the exometer library. To view the metrics, one had to configure an Exometer exporter, which would periodically send the metrics to an external service using the Graphite protocol. Because of the protocol, the metrics would be exported to Graphite or InfluxDB version 1 . One could also query a limited subset of metrics using our GraphQL API (or the legacy REST API) or with the command line interface. Alternatively, metrics could be retrieved from the Erlang shell of a running MongooseIM node.

Logs are another type of instrumentation present in the code. They inform about events occurring in the system and since version 4, they are events with extensible map-like structure and can be formatted e.g. as plain text or JSON. Subsequently, they can be shown in the console or stored in files. You can also set up a log management system like the Elastic (ELK) Stack or Splunk – see the documentation for more details.

The diagram below shows how these two types of instrumentation can work together:

The first observation is that the instrumented code needs to separately call the log and metric API. Updating a metric and logging an event requires two distinct function calls. Moreover, if there are multiple metrics (e.g. execution time and total number of calls), there would be multiple function calls required. There is potential for inconsistency between metrics, or between metrics and logs, because an error could happen between the function calls. The main issue of this solution is however the hardcoding of Exometer as the metric library and the limitation of the Graphite protocol used to push the metrics to external services.

Instrumentation rework in MongooseIM 6.3

The lack of support for the modern and widespread Prometheus protocol was one of the main reasons for the complete rework of instrumentation in version 6.3. Let’s see the updated diagram of MongooseIM instrumentation:

The most noticeable difference is that in the instrumented code, there is just one event emitted. Such an event is identified by its name and a key-value map of labels and contains measurements (with optional metadata) organised in a key-value map. Each event has to be registered before its instances are emitted with particular measurements. The point of this preliminary step is not only to ensure that all events are handled but also to provide additional information about the event, e.g. the measurement keys that will be used to update metrics. Emitted events are then handled by configurable handlers . Currently, there are three such handlers. Exometer and Logger work similarly as before, but there is a new Prometheus handler as well, which stores the metrics internally in a format compatible with Prometheus and exposes them over an HTTP API. This means that any external service can now scrape the metrics using the Prometheus protocol. The primary case would be to use Prometheus for metrics collection, and a graphical tool like Grafana for display. If you however prefer InfluxDB version 2, you can easily configure a scraper , which would periodically put new data into InfluxDB.

As you can see in the diagram, logs can be also emitted directly, bypassing the instrumentation API. This is the case for multiple logs in the system, because often there is no need for any metrics, and a log message is enough. In the future though, we might decide to fully replace logs with instrumentation events, because they are more extensible.

Apart from supporting the Prometheus protocol, additional benefits of the new solution include easier configuration, extensibility, and the ability to add more handlers in the future. You can also have multiple handlers enabled simultaneously, allowing you to gradually change your metric backend from Exometer to Prometheus. Conversely, you can also disable all instrumentation, which was not possible prior to version 6.3. Although it might make little sense at first glance, because it can render the system a black box, it can be useful to gain extra performance in some cases, e.g. if the external metrics like CPU usage are enough, in case of an isolated embedded system, or if resources are very limited.

The table below compares the legacy metrics solution with the new instrumentation framework:

There are about 140 events in total, and some of them have multiple dimensions. You can find an overview in the documentation . In terms of dashboards for tools like Grafana, we believe that each use case of MongooseIM deserves its own. If you are interested in getting one tailored to your needs, don’t hesitate to contact us .

Using the instrumentation

Let’s see the new instrumentation in action now. Starting with configuration, let’s examine the new additions to the default configuration file :

[[listen.http]]
  port = 9090
  transport.num_acceptors = 10

  [[listen.http.handlers.mongoose_prometheus_handler]]
    host = "_"
    path = "/metrics"

(...)

[instrumentation.prometheus]

[instrumentation.log]

The first section, [[listen.http]] , specifies the Prometheus HTTP endpoint. The following [instrumentation.*] sections enable the Prometheus and Log handlers with the default settings – in general, instrumentation events are logged on the DEBUG level, but you can change it. This configuration is all you need to see the metrics at http://localhost:9091/metrics when you start MongooseIM.

As a second example, let’s say that you want only the Graphite protocol integration. In this case, you might configure MongooseIM to use only the Exometer handler, which would push the metrics prefixed with mim to the influxdb1 host every 60 seconds:

[[instrumentation.exometer.report.graphite]]
  interval = 60_000
  prefix = "mim"
  host = "influxdb1"

There are more options possible, and you can find them in the documentation .

Tracing – ad-hoc instrumentation

There is one more type of observability available in Erlang systems, which is tracing . It enables a user to have a more in-depth look into the Erlang processes, including the functions being called and the internal messages being exchanged. It is meant to be used by Erlang developers, and should not be used in production environments because of the impact it can have on a running system. It is good to know, however, because it could be helpful to diagnose unusual issues. To make tracing more user-friendly, MongooseIM now includes erlang_doctor with some MongooseIM-specific utilities (see the tr_util module). This tool provides low-level ad-hoc instrumentation, allowing you to instrument functions in a running system, and gather the resulting data in an in-memory table, which can be then queried, processed, and – if needed – exported to a file. Think of it as a backup solution, which could help you diagnose hidden issues, should you ever experience one.

CockroachDB – a database that scales with MongooseIM

MongooseIM works best when paired with a relational database like PostgreSQL or MySQL, enabling easy cluster node discovery with CETS and persistent storage for users’ accounts, archived messages and other kinds of data. Although such databases are not horizontally scalable out of the box, you can use managed solutions like Amazon Aurora , AlloyDB or Azure Cosmos DB for PostgreSQL . The downsides are the possible vendor lock-in and the fact that you cannot host and manage the DB yourself. With version 6.3 however, the possibilities are extended to CockroachDB . This PostgreSQL-compatible distributed database can be used either as a provider-independent cloud-based solution or as an internally hosted cluster. You can instantly set it up in your local environment and take advantage of the horizontal scalability of both MongooseIM and CockroachDB. If you want to learn how to deploy both MongooseIM and CockroachDB in Kubernetes, see the documentation for CockroachDB and the Helm chart for MongooseIM, together with our recent blog post about setting up an auto-scalable cluster. If you are interested in having an auto-scalable solution deployed for you, please consider our MongooseIM Autoscaler .

Summary

MongooseIM 6.3.0 opens new possibilities for observability – the Prometheus protocol is supported instantly with a new reworked instrumentation layer underneath, guaranteeing ease of future extensions. Regarding database integration, you can now use CockroachDB to store all your persistent data. Apart from these changes, the latest version introduces a multitude of improvements and updates – see the release notes for more information. As the next step, we recommend visiting our product page to see the possible options of support and the services we offer. You can also try the server out at trymongoose.im . In any case, should you have any further questions, feel free to contact us .

The post MongooseIM 6.3: Prometheus, CockroachDB and more appeared first on Erlang Solutions .

This blog post will guide you through the process of setting up an ejabberd Community Server using Docker and Docker Compose , and will also introduce Watchtower for automatic updates. This approach ensures that your configuration remains secure and up to date.

Furthermore, we will examine the potential risks associated with automatic updates and suggest Diun as an alternative tool for notification-based updates.

1. Prerequisites

Please ensure that Docker and Docker Compose are installed on your system.
It would be beneficial to have a basic understanding of Docker concepts, including containers, volumes, and bind-mounts.

2. Set up ejabberd in a docker container

Let’s first create a minimal Docker Compose configuration to start an ejabberd instance.

2.1: Prepare the directories

For this setup, we will create a directory structure to store the configuration, database, and logs. This will assist in maintaining an organised setup, facilitating data management and backup.

mkdir ejabberd-setup && cd ejabberd-setup
touch docker-compose.yml
mkdir conf
touch conf/ejabberd.yml
mkdir database
mkdir logs

This should give you the following structure:

ejabberd-setup/
├── conf
│   └── ejabberd.yml
├── database
├── docker-compose.yml
└── logs

To verify the structure, use the tree command. It is a very useful tool which we use on a daily basis.

Set permissions

Since we&aposll be using bind mounts in this example, it’s important to ensure that specific directories (like database and logs) have the correct permissions for the ejabberd user inside the container (UID 9000 , GID 9000 ).

Customize or skip depending on your needs:

sudo chown -R 9000:9000 database
sudo chown -R 9000:9000 logs

Based on this Issue .

2.2: The docker-compose.yml file

Now, create a docker-compose.yml file inside, containing:

services:
  ejabberd:
    image: ejabberd/ecs:latest
    container_name: ejabberd
    ports:
      - "5222:5222"  # XMPP Client
      - "5280:5280"  # Web Admin Interface, optional
    volumes:
      - ./database:/home/ejabberd/database
      - ./ejabberd.yml:/home/ejabberd/conf/ejabberd.yml
      - ./logs:/home/ejabberd/logs
    restart: unless-stopped

2.3: The ejabberd.yml file

A basic configuration file for ejabberd will be required. we will name it conf/ejabberd.yml .

loglevel: 4
hosts:
- "localhost"

acl:
  admin:
    user:
      - "admin@localhost"

access_rules:
  local:
    allow: all

listen
  -
    port: 5222
    module: ejabberd_c2s

  -
    port: 5280                       # optional
    module: ejabberd_http            # optional
    request_handlers:                # optional
      "/admin": ejabberd_web_admin   # optional

Did you know? Since 23.10 , ejabberd now offers users the option to create or update the relevant MySQL, PostgreSQL or SQLite tables automatically with each update. You can read more about it here .

3: Starting ejabberd

Finally, we&aposre set: you can run the following command to start your stack: docker-compose up -d

Your ejabberd instance should now running in a Docker container! Good job! 🎉

From there, customize ejabberd to your liking! Naturally, in this example we&aposre going to keep ejabberd in its barebones configuration, but we recommend that you configure it as you wish at this stage, to suit your needs (Domains, SSL, favorite modules, chosen database, admin accounts, etc.)

Example: You could register your admin account at this stage

To use the admin interface, you need to create an admin account. You can do so by running the following command:

$ docker exec -it ejabberd bin/ejabberdctl register admin localhost very_secret_password
> User admin@localhost successfully registered

Once this step is complete, you will then be able to access the web admin interface at http://localhost:5280/admin .

4. Set up automatic updates

Finally, we come to the most interesting part: how do I keep my containers up to date?

To keep your ejabberd instance up-to-date, you can use Watchtower , a Docker container that automatically updates other containers when new versions are available.

Warning: Auto-updates are undoubtedly convenient, but they can occasionally cause issues if an update includes breaking changes. Always test updates in a staging environment and back up your data before enabling auto-updates. Further information can be found at the end of this post.

If greater control over updates is required (for example, for mission-critical production servers or clusters), we recommend using Diun , which can notify you of available updates and allow you to decide when to apply them.

4.1: Add Watchtower to your docker-compose.yml

To include Watchtower , add it as a service in docker-compose.yml :

services:
  ejabberd:
    image: ejabberd/ecs:latest
    container_name: ejabberd
    ports:
      - "5222:5222"  # XMPP Client
      - "5280:5280"  # Web Admin Interface, optional
    volumes:
      - ./database:/home/ejabberd/database
      - ./ejabberd.yml:/home/ejabberd/conf/ejabberd.yml
      - ./logs:/home/ejabberd/logs
    restart: unless-stopped

  watchtower:
    image: containrrr/watchtower
    container_name: watchtower
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock
    environment:
      - WATCHTOWER_POLL_INTERVAL=3600 # Sets how often Watchtower checks for updates (in seconds).
      - WATCHTOWER_CLEANUP=true # Ensures old images are cleaned up after updating.
    restart: unless-stopped

Watchtower offers a wide range of additional features, including the ability to set up notifications, exclude specific containers, and more. For further information, please refer to the Watchtower Docs .

Once the docker-compose.yml has been updated, please bring it up using the following command: docker-compose up -d

And.... here you go, you&aposre all set!

5. Best Practices & closing words

Now Watchtower will now perform periodic checks for updates to your ejabberd container and apply them automatically.

Well to be fair, by default if other containers are running on the same server, Watchtower will also update them. This behaviour can be controlled with the help of environment variables (see Container Selection ), which will assist in excluding containers from updates.

One important thing to understand is that Watchtower will only update containers tagged with the :latest tag.

In an environment with numerous Docker containers, using the latest tag streamlines the process of automatic updates. However, it may introduce unanticipated changes with each new, disruptive update. Ideally, we recommend always setting a speficic version like ejabberd/ecs:24.10 and deciding how/when to update it manually (especially if you&aposre into infra-as-code ).

However, we recognise that some users may prefer the convenience of automatic updates, personnally that&aposs what I do my homelab but I&aposm not scared to dig in if stuff breaks.

tl;dr: For a small community server/homelab/personnal instance, Watchtower will help keep things up to date with minimal effort. However, for bigger production environments, it is advisable to tag specific versions to ensure greater control and resilience and update them manually.

With this setup, you now have a fully functioning XMPP server using ejabberd, with automatic updates. You can now start building your chat applications or integrate it with your existing services! 🚀

In the first part of this blog post , I explained how the Matrix protocol works, contrasted its design philosophy with XMPP, and discussed why these differences lead to performance costs in Matrix. Matrix processes each conversation as a graph of events, merged in real-time ^[1] .

Merge operations can be costly in Matrix for large rooms, affecting both database storage and load and disk usage when memory is exhausted, reaching swap level .

That said, there is still room for improvement in the protocol. We have designed and tested slight changes that could make Matrix much more efficient for large rooms.

A Proposal to Simplify and Speed Up Merge Operations

Here is the rationale behind a proposal we have made to simplify and speed up merge operations:

State resolution v2 uses certain graph algorithms, which can result in at least linear processing time for the number of state events in a room’s DAG, creating a significant load on servers.

The goal of this issue is to discuss and develop changes to state resolution to achieve O(n log ⁡ n) total processing time when handling a room with n state events (i.e., O(log ⁡ n) on average) in realistic scenarios, while maintaining a good user experience.

The approach described below is closer to state resolution v1 but seeks to address state resets in a different way.

For more detail, you can read our proposal on the Matrix spec tracker: Make state resolution faster .

In simpler terms, we propose adding a version associated with each event_id to simplify conflict management and introduce a heuristic that skips traversal of large parts of the graph.

Impact of the Proposal

From our initial assessment, in a very large room — such as one with 100,000 members — our approach could improve processing performance by 100x to 1000x, as the current processing cost scales with the number of users in the room. This improvement would enable smoother conversations, reduced lag, and more responsive interactions for end-users, while also reducing server infrastructure load and resource usage.

While our primary goal is to improve performance in very large rooms, these changes benefit all users by reducing overall server load and improving processing times across various room sizes.

We plan to implement this improvement in our own code to evaluate its real-world effectiveness while the Matrix team considers its potential value for the reference protocol.

1. For those who remember, a conversation in Matrix is similar to the collaborative editing protocol built on top of XMPP for the Google Wave platform.

The Ignite Realtime community is happy to be able to announce the immediate availability of version 4.9.1 of Openfire , its cross-platform real-time collaboration server based on the XMPP protocol!

4.9.1 is a bugfix and maintenance release. Among its most important fixes is one for a memory leak that affected all recent versions of Openfire (but was likely noticeable only on those servers that see high volume of users logging in and out). The complete list of changes that have gone into this release can be seen in the change log .

Please give this version a try! You can download installers of Openfire here . Our documentation contains an upgrade guide that helps you update from an older version.

The integrity of these artifacts can be checked with the following sha256sum values:

8c489503f24e35003e2930873037950a4a08bc276be1338b6a0928db0f0eb37d  openfire-4.9.1-1.noarch.rpm
1e80a119c4e1d0b57d79aa83cbdbccf138a1dc8a4086ac10ae851dec4f78742d  openfire_4.9.1_all.deb
69a946dacd5e4f515aa4d935c05978b5a60279119379bcfe0df477023e7a6f05  openfire_4_9_1.dmg
c4d7b15ab6814086ce5e8a1d6b243a442b8743a21282a1a4c5b7d615f9e52638  openfire_4_9_1.exe
d9f0dd50600ee726802bba8bc8415bf9f0f427be54933e6c987cef7cca012bb4  openfire_4_9_1.tar.gz
de45aaf1ad01235f2b812db5127af7d3dc4bc63984a9e4852f1f3d5332df7659  openfire_4_9_1_x64.exe
89b61cbdab265981fad4ab4562066222a2c3a9a68f83b6597ab2cb5609b2b1d7  openfire_4_9_1.zip

We would love to hear from you! If you have any questions, please stop by our community forum or our live groupchat . We are always looking for volunteers interested in helping out with Openfire development!

For other release announcements and news follow us on Mastodon or X

1 post - 1 participant

Read full topic

Adopting machine learning for business is necessary for companies that want to sharpen their competitive industries. With the global market for machine learning projected to reach an impressive $210 billion by 2030 , businesses are keen to seek active solutions that streamline processes and improve customer interactions.

While organisations may already employ some form of data analysis, traditional methods can need more sophistication to address the complexities of today’s market. Businesses that consider optimising machines unlock valuable data insights, make accurate predictions and deliver personalised experiences that truly resonate with customers, ultimately driving growth and efficiency.

What is Machine Learning?

Machine learning (ML) is a subset of artificial intelligence (AI). It uses machine learning algorithms , designed to learn from data, identify patterns, and make predictions or decisions, without explicit programming. By analysing patterns in the data, a machine learning algorithm identifies key features that define a particular data point, allowing it to apply this knowledge to new, unseen information.

Fundamentally data-driven, machine learning relies on vast information to learn, adapt, and improve over time. Its predictive capabilities allow models to forecast future outcomes based on the patterns they uncover. These models are generalisable, so they can apply insights from existing data to make decisions or predictions in unfamiliar situations.

You can read more about machine learning and AI in our previous post .

Approaches to Machine Learning

Machine learning for business typically involves two key approaches: supervised and unsupervised learning , each suited to different types of problems. Below, we explain each approach and provide examples of machine learning use cases where these techniques are applied effectively.

• Supervised Machine Learning: This approach demands labelled data, where the input is matched with the correct output. The algorithms learn to map inputs to outputs based on this training set, honing their accuracy over time.
• Unsupervised Machine Learning: In contrast, unsupervised learning tackles unlabelled data, compelling the algorithm to uncover patterns and structures independently. This method can involve tasks like clustering and dimensionality reduction. While unsupervised techniques are powerful, interpreting their results can be tricky, leading to challenges in assessing whether the model is truly on the right track.

Example of Supervised vs unsupervised learning

Supervised learning uses historical data to make predictions, helping businesses optimise performance based on past outcomes. For example, a retailer might use supervised learning to predict customer churn . By feeding the algorithm data such as customer purchase history and engagement metrics, it learns to identify patterns that indicate a high risk of churn, allowing the business to implement proactive retention strategies.

Unsupervised learning , on the other hand, uncovers hidden patterns within data. It is particularly useful for discovering new customer segments without prior labels. For instance, an e-commerce platform might use unsupervised learning to group customers by their browsing habits, discovering niche audiences that were previously overlooked.

The Impact of Machine Learning on Business

A recent survey by McKinsey revealed that 56% of organisations surveyed are using machine learning in at least one business function to optimise their operations. This growing trend shows how machine learning for business is becoming integral to staying competitive.

The AI market as a whole is also on an impressive growth trajectory, projected to reach USD 407.0 billion by 2027 .

AI Global Market Forecast to 2030

We’re expected to see an astounding compound growth rate (CAGR) of 35.7% by 2030, proving that business analytics is no longer just a trend; it’s moving into a core component of modern enterprises.

Machine Learning for Business Use Cases

Machine learning can be used in numerous ways across industries to enhance workflows. From image recognition to fraud detection , businesses are actively using AI to streamline operations.

Image Recognition

Image recognition, or image classification is a powerful machine learning technique used to identify and classify objects or features in digital images.

Artificial intelligence (AI) and machine learning (ML) are revolutionising image recognition systems by uncovering hidden patterns in images that may not be visible to the human eye. This technology allows these systems to make independent and informed decisions, significantly reducing the reliance on human input and feedback.

As a result, visual data streams can be processed automatically at an ever-increasing scale, streamlining operations and enhancing efficiency. By harnessing the power of AI, businesses can leverage these insights to improve their decision-making processes and gain a competitive edge in their respective markets.

It plays a crucial role in tasks like pattern recognition, face detection, and facial recognition, making it indispensable in security and social media sectors.

Fraud Detection

With financial institutions handling millions of transactions daily, distinguishing between legitimate and fraudulent activity can be a challenge. As online banking and cashless payments grow, so too has the volume of fraud. A 2023 report from TransUnion revealed a 122% increase in digital fraud attempts in the US between 2019 and 2022.

Machine learning helps businesses by flagging suspicious transactions in real-time, with companies like Mastercard using AI to predict and prevent fraud before it occurs, protecting consumers from potential theft.

Speech Recognition

Voice commands have become a common feature in smart devices, from setting timers to searching for shows.

Thanks to machine learning, devices like Google Nest speakers and Amazon Blink security systems can recognise and act on voice inputs, making hands-free operation more convenient for users in everyday situations.

Improved Healthcare

Machine learning in healthcare has led to major improvements in patient care and medical discoveries. By analysing vast amounts of healthcare data, machine learning enhances the accuracy of diagnoses, optimises treatments, and accelerates research outcomes.

For instance, AI systems are already employed in radiology to detect diseases in medical images, such as identifying cancerous growths. Additionally, machine learning is playing a crucial role in genomic research by uncovering patterns linked to genetic disorders and potential therapies. These advancements are paving the way for improved diagnostics and faster medical research, offering tremendous potential for the future of healthcare.

Key applications of machine learning in healthcare include:

• Developing predictive modelling
• Improving diagnostic accuracy
• Personalising patient care
• Automating clinical workflows
• Enhancing patient interaction

Machine learning in healthcare utilises algorithms and statistical models to analyse large medical datasets, facilitating better decision-making and personalised care. As a subset of AI, machine learning identifies patterns, makes predictions, and continuously improves by learning from data. Different types of learning, including supervised and unsupervised learning, find applications in disease classification and personalised treatment recommendations.

Chatbots

Many businesses rely on customer support to maintain satisfaction. However, staffing trained specialists can be expensive and inefficient. AI-powered chatbots, equipped with natural language processing (NLP), assist by handling basic customer queries. This frees up human agents to focus on more complicated issues. Companies can provide more efficient and effective support without overburdening their teams.

Each of these applications offers businesses the chance to streamline operations and improve customer experiences.

Machine Learning Case Studies

Machine learning for business is transforming industries by enabling companies to enhance their operations, improve customer experiences, and drive innovation.

Here are a few machine learning case studies showing how leading organisations have integrated machine learning into their business strategies.

PayPal

PayPal, a worldwide payment platform, faced huge challenges in identifying and preventing fraudulent transactions.

To tackle this issue, the company implemented machine learning algorithms designed for fraud detection . These algorithms analyse various aspects of each transaction, including the transaction location, the device used, and the user’s historical behaviour. This approach has significantly enhanced PayPal’s ability to protect users and maintain the integrity of its payment platform.

YouTube

YouTube has long employed machine learning to optimise its operations, particularly through its recommendation algorithms . By analysing vast amounts of historical data, YouTube suggests videos to its viewers based on their preferences. Currently, the platform processes over 80 billion data points for each user, requiring large-scale neural networks that have been in use since 2008 to effectively manage this immense dataset.

Dell

Recognising the importance of data in marketing, Dell’s marketing team sought a data-driven solution to enhance response rates and understand the effectiveness of various words and phrases. Dell partnered with Persado, a firm that leverages AI to create compelling marketing content. This collab led to an overhaul of Dell’s email marketing strategy, resulting in a 22% average increase in page visits and a 50% boost in click-through rates (CTR). Dell now utilises machine learning methods to refine its marketing strategies across emails, banners, direct mail, Facebook ads, and radio content.

Tesla

Tesla employs machine learning to enhance the performance and features of its electric vehicles. A key application is its Autopilot system , which combines cameras, sensors, and machine learning algorithms to provide advanced driver assistance features such as lane centring, adaptive cruise control, and automatic emergency braking.

The Autopilot system uses deep neural networks to process vast amounts of real-world driving data, enabling it to predict driving behaviour and identify potential hazards. Additionally, Tesla leverages machine learning in its battery management systems to optimise battery performance and longevity by predicting behaviour under various conditions.

Netflix

Netflix is a leader in personalised content recommendations. It uses machine learning to analyse user viewing habits and suggest shows and movies tailored to individual preferences. This feature has proven essential for improving customer satisfaction and increasing subscription renewals. To develop this system, Netflix utilises viewing data—including viewing durations, metadata, release dates, timestamps etc. Netflix then employs collaborative filtering, matrix factorisation, and deep learning techniques to accurately predict user preferences.

Benefits of Machine Learning in Business

If you’re still contemplating the value of machine learning for your business, consider the following key benefits:

As organisations increasingly adopt machine learning, they position themselves to not only meet current demands but poise them for future innovation.

Elixir and Erlang in Machine Learning

As organisations explore machine learning tools, many are turning to Erlang and Elixir programming languages to develop customised solutions that cater to their needs. Erlang’s fault tolerance and scalability make it ideal for AI applications, as described in our blog on adopting AI and machine learning for business . Additionally, Elixir’s concurrency features and simplicity enable businesses to build high-performance AI applications.

Learn more about how to build a machine-learning project in Elixir here .

As organisations become more familiar with AI and machine learning tools, many are turning to Erlang and Elixir programming languages to develop customised solutions that cater to their needs.

Elixir, built on the Erlang virtual machine (BEAM), delivers top concurrency and low latency. Designed for real-time, distributed systems, Erlang prioritises fault tolerance and scalability, and Elixir builds on this foundation with a high-level, functional programming approach. By using pure functions and immutable data, Elixir reduces complexity and minimises unexpected behaviours in code. It excels at handling multiple tasks simultaneously, making it ideal for AI applications that need to process large amounts of data without compromising performance.

Elixir’s simplicity in problem-solving also aligns perfectly with AI development, where reliable and straightforward algorithms are essential for machine learning. Furthermore, its distribution features make deploying AI applications across multiple machines easier, meeting the high computational demands of AI systems.

With a rich ecosystem of libraries and tools, Elixir streamlines development, so AI applications are scalable, efficient, and reliable. As AI and machine learning become increasingly vital to business success, creating high-performing solutions will become a key competitive advantage.

Final Thoughts

Embracing machine learning for business is no longer optional for companies that want to remain competitive. Machine learning tools empower businesses to make faster, data-driven decisions, streamline operations, and offer personalised customer experiences. Contact the Erlang Solutions team today if you’d like to discuss building AI systems using Elixir and Erlang or for more insights into implementing machine learning solutions,

The post Why you should consider machine learning for business appeared first on Erlang Solutions .

We’re excited to announce ejabberd 24.10, a major release packed with substantial improvements and support for important extensions specified by the XMPP Standard Foundation (XSF). This release represents three months of focused development, bringing around 100 commits to the core repository alongside key updates in dependencies. The improvements span enhanced security, streamlined connectivity, and new administrative tools—all designed to make ejabberd more powerful and easier to use than ever.

Release Highlights:

• XEP-0288: Bidirectional Server-to-Server Connections
• XEP-0480: SASL Upgrade Tasks
• IQ permission in privileged entities
• PubSub varied fixes
• WebAdmin improvements

If you are upgrading from a previous version, please note minor changes in commands and two changes in hooks . There are no configuration or SQL schema changes in this release.

Below is a detailed breakdown of the new features, fixes, and enhancements:

Support for XEP-0288: Bidirectional Server-to-Server Connections

The new mod_s2s_bidi module introduces support for XEP-0288: Bidirectional Server-to-Server Connections . This update removes the requirement for two connections per server pair in XMPP federations, allowing for more streamlined inter-server communications. However, for full compatibility, ejabberd can still connect to servers that do not support bidirectional connections, using two connections when necessary. The module is enabled by default in the sample configuration.

Support for XEP-0480: SASL Upgrade Tasks

The new mod_scram_upgrade module implements XEP-0480: SASL Upgrade Tasks . Compatible clients can now automatically upgrade encrypted passwords to more secure formats, enhancing security with minimal user intervention.

PubSub Service Improvements

We’ve implemented six noteworthy fixes to improve PubSub functionality:

• PEP notifications are sent only to owners when +notify ( 3469a51 )
• Non-delivery errors for locally generated notifications are now skipped ( d4b3095 )
• Fix default node config parsing ( b439929 )
• Fix merging of default node options ( ca54f81 )
• Fix choice of node config defaults ( a9583b4 )
• Fall back to default plugin options ( 36187e0 )

IQ permission for privileged entities

The mod_privilege module now supports IQ permission based on version 0.4 of XEP-0356: Privileged Entity . See #3889 for details. This feature is especially useful for XMPP gateways using the Slidge library.

WebAdmin improvements

ejabberd 24.06 release laid the foundation for a more streamlined WebAdmin interface, reusing existing commands instead of using specific code, with a possibly different logic. This major change allows developers to add new pages very fast, just by calling existing commands. It also allows administrators to use the same commands than in ejabberdctl or any other command frontend .

As a result, many new pages and content were added. Building on that, the 24.10 update introduces MAM (Message Archive Management) support, allowing administrators to view message counts, remove all MAM messages, or only for a specific contact, and also view the MAM Archive directly from WebAdmin.

Additionally, WebAdmin now hides pages related to modules that are disabled, preventing unnecessary options from displaying. This affects mod_last, mod_mam, mod_offline, mod_privacy, mod_private, mod_roster, mod_vcard.

Fixes in commands

• set_presence : Now returns an error when the session is not found.

• send_direct_invitation : Improved handling of malformed JIDs.

• update : Fix command output. So far, ejabberd_update:update/0 returned the return value of release_handler_1:eval_script/1 . That function returns the list of updated but unpurged modules, i.e., modules where one or more processes are still running an old version of the code. Since commit 5a34020d23f455f80a144bcb0d8ee94770c0dbb1 , the ejabberd update command assumes that value to be the list of updated modules instead. As that seems more useful, modify ejabberd_update:update/0 accordingly. This fixes the update command output.

• get_mam_count : New command to retrieve the number of archived messages for a specific account.

Changes in hooks

Two key changes in hooks:

• New check_register_user hook in ejabberd_auth.erl to allow blocking account registration when a tombstone exists.

• Modified room_destroyed hook in mod_muc_room.erl . Until now the hook passed as arguments: LServer, Room, Host . Now it passes: LServer, Room, Host, Persistent That new Persistent argument passes the room persistent option, required by mod_tombstones because only persistent rooms should generate a tombstone, temporary ones should not. And the persistent option should not be completely overwritten, as we must still known its real value even when room is being destroyed.

Log Erlang/OTP and Elixir versions

During server start, ejabberd now shows in the log not only its version number, but also the Erlang/OTP and Elixir versions being used. This will help the administrator to determine what software versions are being used, which is specially useful when investigating some problem, and explaining it to other people for help.

The ejabberd.log file now looks like this:

...
2024-10-22 13:47:05.424 [info] Creating Mnesia disc_only table &aposoauth_token&apos
2024-10-22 13:47:05.427 [info] Creating Mnesia disc table &aposoauth_client&apos
2024-10-22 13:47:05.455 [info] Waiting for Mnesia synchronization to complete
2024-10-22 13:47:05.591 [info] ejabberd 24.10 is started in the node :ejabberd@localhost in 1.93s
2024-10-22 13:47:05.606 [info] Elixir 1.16.3 (compiled with Erlang/OTP 26)
2024-10-22 13:47:05.606 [info] Erlang/OTP 26 [erts-14.2.5.4] [source] [64-bit] [smp:4:4] [ds:4:4:10] [async-threads:1] [jit:ns]

2024-10-22 13:47:05.608 [info] Start accepting TCP connections at 127.0.0.1:7777 for :mod_proxy65_stream
2024-10-22 13:47:05.608 [info] Start accepting UDP connections at [::]:3478 for :ejabberd_stun
2024-10-22 13:47:05.608 [info] Start accepting TCP connections at [::]:1883 for :mod_mqtt
2024-10-22 13:47:05.608 [info] Start accepting TCP connections at [::]:5280 for :ejabberd_http
...

Brand new ProcessOne and ejabberd web sites

We’re excited to unveil the redesigned ProcessOne website, crafted to better showcase our expertise in large-scale messaging across XMPP, MQTT, Matrix, and more. This update highlights our core mission of delivering scalable, reliable messaging solutions, with a fresh layout and streamlined structure that reflect our cutting-edge work in the field.

You now get a cleaner ejabberd page , offering quick access to important URLs for downloads, blog posts, and documentation.

Behind the scenes, we’ve transitioned from WordPress to Ghost, a move inspired by its efficient, user-friendly authoring tools and long-term maintainability. All previous blog content has been preserved, and with this new setup, we’re poised to deliver more frequent updates on messaging, XMPP, ejabberd, and related topics.

We welcome your feedback—join us on our new site to share your thoughts, or let us know about any issue or broken link!

Acknowledgments

We would like to thank the contributions to the source code, documentation, and translation provided for this release by:

• Holger Weiß for PubSub and other fixes
• Michael Slezak for mix release fix
• Guus der Kinderen for the XMPP Interop tests
• Sketch6580 , updated the Chinese translation
• Wellington Uemura , updated the Portuguese (Brazil) translation
• Ermete Melchiorre , updated the Italian translation
• Mr. EddX , updated the Bulgarian translation

And also to all the people contributing in the ejabberd chatroom, issue tracker...

Improvements in ejabberd Business Edition

Customers of the ejabberd Business Edition , in addition to all those improvements and bugfixes, also get MUC support in mod_unread .

ejabberd keeps a counter of unread messages per conversation using the mod_unread module. This now also works in MUC rooms: each user can retrieve the number of unread messages in each of their rooms.

ChangeLog

This is a more detailed list of changes in this ejabberd release:

Miscelanea

• ejabberd_c2s : Optionally allow unencrypted SASL2
• ejabberd_system_monitor : Handle call by gen_event:swap_handler ( #4233 )
• ejabberd_http_ws : Remove support for old websocket connection protocol
• ejabberd_stun : Omit auth_realm log message
• ext_mod : Handle info message when contrib module transfers table ownership
• mod_block_strangers : Add feature announcement to disco-info ( #4039 )
• mod_mam : Advertise XEP-0424 feature in server disco-info ( #3340 )
• mod_muc_admin : Better handling of malformed jids in send_direct_invitation command
• mod_muc_rtbl : Fix call to gen_server:stop ( #4260 )
• mod_privilege : Support "IQ permission" from XEP-0356 0.4.1 ( #3889 )
• mod_pubsub : Don&apost blindly echo PEP notification
• mod_pubsub : Skip non-delivery errors for local pubsub generated notifications
• mod_pubsub : Fall back to default plugin options
• mod_pubsub : Fix choice of node config defaults
• mod_pubsub : Fix merging of default node options
• mod_pubsub : Fix default node config parsing
• mod_register : Support to block IPs in a vhost using append_host_config ( #4038 )
• mod_s2s_bidi : Add support for S2S Bidirectional
• mod_scram_upgrade : Add support for SCRAM upgrade tasks
• mod_vcard : Return error stanza when storage doesn&apost support vcard update ( #4266 )
• mod_vcard : Return explicit error stanza when user attempts to modify other&aposs vcard
• Minor improvements to support mod_tombstones (#2456)
• Update fast_xml to use use_maps and remove obsolete elixir files
• Update fast_tls and xmpp to improve s2s fallback for invalid direct tls connections
• make-binaries : Bump dependency versions: Elixir 1.17.2, OpenSSL 3.3.2, ...

Administration

• ejabberdctl : If ERLANG_NODE lacks host, add hostname ( #4288 )
• ejabberd_app : At server start, log Erlang and Elixir versions
• MySQL: Fix column type in the schema update of archive table in schema update

Commands API

• get_mam_count : New command to get number of archived messages for an account
• set_presence : Return error when session not found
• update : Fix command output
• Add mam and offline tags to the related purge commands

Code Quality

• Fix warnings about unused macro definitions reported by Erlang LS
• Fix Elvis report: Fix dollar space syntax
• Fix Elvis report: Remove spaces in weird places
• Fix Elvis report: Don&apost use ignored variables
• Fix Elvis report: Remove trailing whitespace characters
• Define the types of options that opt_type.sh cannot derive automatically
• ejabberd_http_ws : Fix dialyzer warnings
• mod_matrix_gw : Remove useless option persist
• mod_privilege : Replace try...catch with a clean alternative

Development Help

• elvis.config : Fix file syntax, set vim mode, disable many tests
• erlang_ls.config : Let it find paths, update to Erlang 26, enable crossref
• hooks_deps : Hide false-positive warnings about gen_mod
• Makefile : Add support for make elvis when using rebar3
• .vscode/launch.json : Experimental support for debugging with Neovim
• CI: Add Elvis tests
• CI: Add XMPP Interop tests
• Runtime: Cache hex.pm archive from rebar3 and mix

Documentation

• Add links in top-level options documentation to their Docs website sections
• Document which SQL servers can really use update_sql_schema
• Improve documentation of ldap_servers and ldap_backups options ( #3977 )
• mod_register : Document behavior when access is set to none ( #4078 )

Elixir

• Handle case when elixir support is enabled but not available
• Start ExSync manually to ensure it&aposs started if (and only if) Relive
• mix.exs : Fix mix release error: logger being regular and included application ( #4265 )
• mix.exs : Remove from extra_applications the apps already defined in deps ( #4265 )

WebAdmin

• Add links in user page to offline and roster pages
• Add new "MAM Archive" page to webadmin
• Improve many pages to handle when modules are disabled
• mod_admin_extra : Move some webadmin pages to their modules

Full Changelog

https://github.com/processone/ejabberd/compare/24.07...24.10

ejabberd 24.10 download & feedback

As usual, the release is tagged in the Git source code repository on GitHub .

The source package and installers are available in ejabberd Downloads page. To check the *.asc signature files, see How to verify ProcessOne downloads integrity .

For convenience, there are alternative download locations like the ejabberd DEB/RPM Packages Repository and the GitHub Release / Tags .

The ecs container image is available in docker.io/ejabberd/ecs and ghcr.io/processone/ecs . The alternative ejabberd container image is available in ghcr.io/processone/ejabberd .

If you consider that you&aposve found a bug, please search or fill a bug report on GitHub Issues .