Does anyone know how to connect/admin cluster via GHCI

[joelmccracken]

I recently tried to connect and struggled. I don't remember specifically what I thought was going wrong, but yeah.

Does anyone know how to, say, fire up a process somewhere, and then send messages to it with GHCi, etc?

[hughjfchen]

I've been exploring Cloud Haskell recently and also want to know how to do that within GHCI. Could you tell me what had you done about this? And what errors did you get?

[joelmccracken]

So, I just wanted to dump my thoughts here before I for get to again.

I recently tried this again and was foiled.

An example of what I wanted to do would be something like: open two ghci sessions, join the same network, start up a Process in each, send messages back and forth.

I think generally this could be done with GHCi today on some sense, but the types just dont exaclty work so that I can interactively "interact" like a process, as you need to be within the Process a type, not at the top of the repl. If I could somehow say "ok, run this, then open a new REPL session from within here", I think that might do it, but getting that to happen with GHCi wont be easy afaict.

I did explore the ghci debugger, but wasn't able to make that work either. I know some enhancements are coming to debugging in recent GHC versions, so maybe that could help. I forget the details.

However, I'm currently wondering if something like this might do: https://hackage.haskell.org/package/hint

[davidsd]

The Process monad is a ReaderT LocalProcess over IO. To use it in ghci, maybe we can expose enough of the LocalProcess type/constructors that the user can create their own and pass it explicitly to each Process computation, e.g. myLocalProcess <- newLocalProcess runReaderT myProcessComputation1 myLocalProcess :: IO () runReaderT myProcessComputation2 myLocalProcess :: IO () ... (It's also possible that enough functionality is already exposed.) Would this satisfy what you want?

…

On Tue, Sep 23, 2025 at 8:32 AM Joel McCracken ***@***.***> wrote: *joelmccracken* left a comment (haskell-distributed/distributed-process#473) <#473 (comment)> So, I just wanted to dump my thoughts here before I for get to *again*. I recently tried this again and was foiled. An example of what I wanted to do would be something like: open two ghci sessions, join the same network, start up a Process in each, send messages back and forth. I think generally this could be done with GHCi today on some sense, but the types just dont exaclty work so that I can interactively "interact" like a process, as you need to be within the Process a type, not at the top of the repl. If I could somehow say "ok, run this, then open a new REPL session from within here", I *think* that might do it, but getting that to happen with GHCi wont be easy afaict. I did explore the ghci debugger, but wasn't able to make that work either. I know some enhancements are coming to debugging in recent GHC versions, so maybe that could help. I forget the details. However, I'm currently wondering if something like this might do: https://hackage.haskell.org/package/hint — Reply to this email directly, view it on GitHub <#473 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABZCIK4X4V4MREQ72GPZPL3UFRXNAVCNFSM6AAAAABY4UMYD2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGMRUGUZTCOBUGI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.*** com>

[joelmccracken]

Yeah, that sounds like what I was coming around to/thinking about; IIRC, I couldn't find out how to do exactly what you're doing there, making a local process and then working with it explicitly.

OK, I went and poked at it, and see this, which looks promising: https://github.com/haskell-distributed/distributed-process/blob/master/packages/distributed-process/src/Control/Distributed/Process/Internal/Types.hs#L337-L339

but then I see: https://github.com/haskell-distributed/distributed-process/blob/master/packages/distributed-process/src/Control/Distributed/Process/Node.hs#L370-L466 which to me to be the only spot that creates local processes. I also don't see a way to get a LocalProcess from a node, well, perhaps excepting the following:

https://github.com/haskell-distributed/distributed-process/blob/master/packages/distributed-process/src/Control/Distributed/Process/Internal/Types.hs#L855-L856

the localProcessWithId and/or more generally the localProcesses bindings seem like perhaps could be used to "get" a LocalProcess,

oh!

https://github.com/haskell-distributed/distributed-process/blob/master/packages/distributed-process/src/Control/Distributed/Process/Internal/Types.hs#L299-L306

Maybe something like this?

• start some process
• use withValidLocalState for the current node to get the list of processes via localProcesses;
• pick one out, use it with runLocalProcess to run it.

I'll have to poke at it again soon. Either way it would be very nice if this was easier to do - or, even, documented!

The main thing that makes me uncertain here about somehow making an easier way to do this is my not knowing the exact details about process lifecycles, guarantees, and/or what might happen if these inner details started to escape in a world where asynchronicity is the norm.