Hello everyone! I`m currently trying to create a demo to learn more about distributed nodes and elixir libraries.
I created this repo and the goal is to create N nodes, make them all see each other, run 1 “cache” (genserver) process across the entire cluster, when the node that is parent of the said process dies, a new process is created inside other node and the last part would be to keep the state when the new process starts.
The cluster part and restart I was able to develop using horde and libcluster libraries but I dont see a way to keep the state across the nodes. Horde says something about handoff and save the state at redis, postgress or something like that… I`d like to develop this using only elixir / erlang. It also mentions delta crdt and other blog posts says something about mnesia but I dont get a way to make neither of them see the dynamic nodes once that all docs I read mentioned to set some pre known node list.
Here is the repo
Can someone please mention an alternative or solution to this problem?
PS: I`m also facing some latency when accessing the endpoint if u spot some problem that could be causing that it would help a lot THANKS 
Can you elaborate on this? Even if the goal is learning, wouldn’t you want to learn good patterns around state persistence rather than bad ones? I suspect it is technically possible to achieve what you’re trying to do but I don’t think it would be a good solution for any real use case, because…
when the node that is parent of the said process dies
…AFAIK there are no guarantees around this that would allow for safe persistence of data.
Assuming you already know all that and want an answer anyway, I would guess that you would want to duplicate each state change across nodes (maybe using something like Phoenix.PubSub) rather than try to “hand it off” just in time.
I`m trying to implement concepts that I saw at this video ElixirConf 2018 - Docker and OTP Friends or Foes - Daniel Azuma Of course the use case is different I didnt develop a game or anything like that I’d just like to use the concepts in something simpler as a cache across nodes.
So I don’t have experience with Horde, but it looks like that works pretty similar to what I described under the hood, viz. the data is replicated across all nodes, so you wouldn’t need anything else, you’d just use Horde to globalize your Genserver process and it will handle the details for you. Hopefully someone with more experience with the library will be able to confirm with more details, but you might want to include “Horde” in your title/tags.
1 Like
