Hot code reloading with Erlang

Hot code reloading is always more work than just blue-green and should be avoided if necessary. For example, author of Learn You Some Erlang writes [1]:

> if you can avoid the whole procedure (which will be called relup from now on) and do simple rolling upgrades by restarting VMs and booting new applications, I would recommend you do so.

Erlang grew out of the challenges faced by telecoms industries such as what do you do when blue-green isn't an option? Think an in-use packet switch that is the only point of contact between two networks. No way to take the switch down for maintenance without some interruption in service, which gets messy when dealing with timeouts. In the Armstrong thesis paper he gives another example [2]:

> Usually in a sequential system, if we wish to change the code, we stop the system, change the code and re-start the program. In certain real-time control systems, we might never be able to turn off the system in order to change the code and so these systems have to be designed so that the code can be changed without stopping the system. An example of such a system is the X2000 satellite control system developed by NASA.

This power comes at a cost, though. LYSE again:

> It is said that divisions of Ericsson that do use relups spend as much time testing them as they do testing their applications themselves. They are a tool to be used when working with products that can imperatively never be shut down.

The point being, hot code reloading is an additional feature that can come in handy but for most of HN's audience probably won't be relevant; the cost outweighs the benefits of just blue-green deploying it.

[1] http://learnyousomeerlang.com/relups#the-hiccups-of-appups-a... [2] http://www.erlang.org/download/armstrong_thesis_2003.pdf

On the contrary, the implementation of Erlang's hot code reloading forces state to be separated from the code. If you look at Erlang's gen_server, every call requires you to return a new State object, which is passed to the next function call.

In other words, you can compare the Erlang's virtual machine with a container itself, and everything old is new again!

Hot code reloading is best used for a scenario where you cannot afford to restart your system, usually because it drags a lot of internal state around and reconstructing that state is expensive.

Typical use cases include several gigabytes of in memory state which takes a long time to read in and get hot when redeploying or a large amount of long-running TCP connections.

For most other uses, we just do rolling upgrades in Erlang as everyone else is doing. It is somewhat simpler to get to work, and immutable architecture is to a certain extent easier to manipulate.

You have to start with the problem. If your problem is solved by highly available and stateful services, then the Erlang VM on the host seems like a good idea. If the availability doesn't matter that much or the services can be made stateless without much pain - go for the containers.

As always: depends on the use case.

We're using Erlang as the primary language environment for our IoT product for a lot of reasons but one big one is: Hot code loading and a very robust release upgrade environment with a lot of control over the process (including restarting everything inside the VM if that's what we wish to do).

For our product, a digital light switch / dimmer, high uptime guarantees is a very important requirement and Erlang has it all plus many other wonderful features.

You can use that to replace code by monkey patching

    $ cat hi.rb 
    def method
      puts "hi"
    end
    method
    load("hello.rb")
    method

    $ cat hello.rb 
    def method
      puts "hello"
    end

    $ ruby hi.rb 
    hi
    hello

You must engineer your application to execute the load method and that's it. However I wonder if this is really equivalent to what Erlang does. I remember http://rvirding.blogspot.it/2008/01/virdings-first-rule-of-p...

When your server has several gigs of state. It's VERY useful on a dev server. Instead of waiting several minutes for reload, I just load in new code manually (typically I change only 1-2 files per reload). If something breaks - hey, it's only dev server. Erlangs other feature - almost everything works alone - helps with this. If something breaks, it breaks only in one place, so most of the time I only need to make small changes and reload once more. Rest of the system does what it needs without any downgrades.

That's what distributed Erlang is for.

What -exactly- do you want to do when you say you want to restart the Erlang VM?

I'm asking because I don't have enough context to know why you want to do what you're asking to do.