feat: flip swap mechanism — MD as primary, RST as override (Phase 1)

This is the first of three phases inverting the per-page swap mechanism
so MD becomes the canonical primary and RST becomes the rare override.

Phase 1 — file renames + conf.py exclude_patterns flip only:
- Rename docs/**/md-<stem>.md to docs/**/<stem>.md (drop md- prefix)
  for all 254 stems previously listed in docs/_swap.txt
- Rename docs/**/<stem>.rst to docs/**/rst-<stem>.rst (add rst- prefix)
  for the same 254 stems
- Repurpose docs/_swap.txt as docs/_rst_overrides.txt; initially empty
  comment-only since no pages need the RST fallback right now
- conf.py exclude_patterns flipped: rst-*.rst is now excluded by default
  instead of md-*.md
- conf.py runtime-artifact references updated to _rst_override_state.json
  and _md_exclude.txt (Phase 2 will rewrite swap_sources.py to produce
  these names; for now no swap script runs because overrides list is empty)

Phase 2 (next commit on this branch) will rewrite scripts/swap_sources.py
with inverted rename direction, delete scripts/import_myst.py + tests, and
update tests/test_swap_sources.py for the new semantics.

Phase 3 will be the cleanup pass and ready-for-review flip.

Generated by robots https://vyos.io

1 files changed, 38 insertions, 0 deletions

diff --git a/docs/installation/virtual/vmware.md b/docs/installation/virtual/vmware.md
new file mode 100644
index 00000000..66278ae9
--- /dev/null
+++ b/docs/installation/virtual/vmware.md
@@ -0,0 +1,38 @@
+---
+lastproofread: '2026-02-02'
+---
+
+(vyosonvmware)=
+
+# Running on VMware ESXi
+
+## ESXi 5.5 or later
+
+`.ova` files are available for supporting users. You can also set up VyOS
+using a generic Linux instance by attaching the bootable ISO file and
+installing using the `install image` command.
+
+:::{note}
+Previous issues have been documented with GRE/IPSEC tunneling
+using the E1000 adapter on VyOS guests. Use the VMXNET3 adapter instead.
+:::
+
+### Memory Contention Considerations
+
+When the underlying ESXi host reaches approximately 92% memory utilization,
+it begins the balloon process to reclaim memory from guest operating systems.
+This creates artificial memory pressure through the `vmmemctl` driver. Because
+VyOS does not have a swap file by default, this pressure cannot move memory
+data to a paging file. Instead, it consumes memory and forces the guest into
+a low memory state with no recovery option. The balloon can expand to 65% of
+guest allocated memory, so a VyOS guest using more than 35% of memory can
+encounter an out-of-memory situation and trigger the kernel `oom_kill`
+process. The `oom_kill` process then terminates memory-hungry processes.
+
+To prevent ballooning, configure VyOS routers in a resource group with
+adequate memory reservations.
+
+### References
+
+<https://muralidba.blogspot.com/2018/03/how-does-linux-out-of-memory-oom-killer.html>
+