Remove: 9999 limit on MB

[jmealo]

Is your feature request related to a problem? Please describe.
It would be nice if we could specify higher values in megabytes (or allow for decimal places in GB).

Describe the solution you'd like
To remove the 9999 limit on MB.

Describe alternatives you've considered
Forking.

Additional context
Kubernetes resource limits aren't going to have a clear boundary on GB, so MB is the best option.

[le0pard]

Hello. I dont get issue to select unit "MB", if you need "MB". Nobody will use "23846 MB", because in this case easy just switch to GB

[jmealo]

@le0pard: Thank you for the quick response.

• On physical hardware, your argument makes sense.
• On Docker/K8s/Cloud, you rarely have memory allocations that align to GB boundaries.

Use case:

• I need to enter things like 11328MB (see table below for why).
• Validation doesn't allow 11.328GB (or any fractional units)
• There's no reason to have the arbitrary 9999 limitation -- removing it didn't break anything AFAICT.

These are some real values that I have to work with. For example, on Azure AKS:

 Node Memory (GB) | Node vCPU | PostgreSQL Memory Limit (Mi) | PostgreSQL CPU Limit (m) | Pooler Memory Limit (Mi) | Pooler CPU Limit (m) |
|-----------------:|----------:|-----------------------------:|-------------------------:|-------------------------:|---------------------:|
|                8 |         2 |                         6352 |                     1170 |                      128 |                  200 |
|               16 |         4 |                        13725 |                     2934 |                      128 |                  200 |
|               32 |         8 |                        28470 |                     6498 |                      128 |                  200 |
|               64 |        16 |                        57961 |                    13626 |                      128 |                  200 |
|               96 |        24 |                        87453 |                    20394 |                      128 |                  200 |
|              128 |        32 |                       116944 |                    27882 |                      128 |                  200 |
|              256 |        64 |                       234909 |                    56394 |                      128 |                  200 |

For folks hosting PostgreSQL on Kubernetes. The allocatable memory is almost never on a hard GB boundary. It varies by node/instance type.

We're already paying a 35% to 15% memory overhead penalty, so the idea of just truncating our MB to full-gigabytes to workaround an arbitrary limitation doesn't really serve users well.

[jmealo]

I'm happy to remove the non-memory max limit increases as well. I just figured for consistency's sake, the removal of arbitrary limits would be an overall improvement for UX and future proofing things.

[le0pard]

resolved by #102