Closed kristopolous closed 4 years ago
staging and prod both assign the same reflect server
so I'm doing ping forwarding now. There's certainly tradeoffs, the most notable is that we're polluting the production server with units that are ahead in code.
It can certainly lead to confusion as to whose loyalty a screen currently has (this can be alleviated with some database flag) but at the same time affords us a global namespace for the port forwarding which means the screen connection tool syntax doesn't need to change.
This may be an excellent or a terrible idea, depending on the day. This approach marries us to an opinion on the matter and we're free to call off this marriage in the future and have 2 totally self-contained systems
That is what we'll probably be doing anyway in order to avoid the 2^16 problem of this approach. I thought there was documentation somewhere of the host:port syntax that I was thinking to alleviate this (essentially not storing the port as an integer but a host/port pair) but I can't find any reference.
Anyway, that's the true solution to both this and the 2^16 problem to be done at a later date and time that's more motivated and sympathetic to the cause.
two machines are allocating and not talking to each other :-(