Add a clear indication of why resize-swap fails

[olegbbtr]

Problem description / Motivation

#887 introduced the ability to resize swap after the VM has already started. This mechanism is meant to be used upon endpoint binding. Upon resizing, the /neonvm/bin/resize-swap is deleted from the VM, to prevent attempts to resize the swap mid-execution. The way it can currently happen is e.g. compute_ctl restarts.

The reason why resizing swap is not allowed after postgres has started is that resizing assumes swapoff, and then swapon. swapoff in the presence of customer workload may lead to OOMs, so is is not allowed.

However, the way this protection is implemented - by removing the implementation script confuses people, and overall changing a set of scripts mid-execution isn't a good idea.

Feature idea(s) / DoD

The error message upon resizing mid-execution should be clear. We might want to have a way to do the resize forcefully.

Implementation ideas

Probably this functionality will be re-implemented as a part of neonvmd, so no more scripts.

[stradig]

Decision: We will wait for neonvm-daemon to become and thing and then tackle it.

[zenithdb-bot-dev[bot]]

This issue was moved to Jira: LKB-2223