add support for monitoring thp, ballooning, zswap, ksm cow

[ktsaou]

Modified the vmstat module of proc module to monitor:

• transparent huge pages (THP, many charts, including allocations, compaction, splitting, swapout, etc)
• memory ballooning (when Linux runs as guest VM on a host that supports memory ballooning)
• zswap operations
• KSM copy-on-write operations (COW)

Especially the THP is very important since it seems enabled on most Linux distros by default.

[ilyam8]

lgtm, but a small note:

Usually having "events/s" as units is a red flag because it means we are aggregating (chart level) non-aggregatable metrics - grouping by host/system w/o selecting a dimension (the Cloud UI) will give useless values. E.g.

thp_swpout
is incremented every time a huge page is swapout in one
piece without splitting.

thp_swpout_fallback
is incremented if a huge page has to be split before swapout.
Usually because failed to allocate some continuous swap space
for the huge page.

The sum of these events doesn't seem useful/correct to me.

---

And this PR is 15000th contribution to this repo 🎉

[ktsaou]

Usually having "events/s" as units is a red flag because it means we are aggregating (chart level) non-aggregatable metrics - grouping by host/system w/o selecting a dimension (the Cloud UI) will give useless values

How do you suggest to do it?
I am not sure why they are useless when grouped. Why they are?

[ilyam8]

I propose to go with the current implementation, which is why I approved the PR. We do such aggregations pretty often. By "such aggregations" I mean correct when grouped by dimension, wrong when grouped by anything else without selecting a dimension (some obvious examples: system load average, system pressure).

cc @ralphm

Why they are?

I don't find them useful because the sum of:

• huge page swapout events
• huge page split events

doesn't seem to make sense to me, they look like 2 completely different metrics. But my understanding can be wrong because it is based only on metrics description, I am not a specialist.

I am not a specialist

If thp_swpout_fallback means a huge page swapout event and not only split, and this swapout is not accounted in thp_swpout then all good.

[ktsaou]

@cakrit this needs to go into the release notes. The description is in the document I shared with you about hugepages.