Log ru_maxrss from getrusage for PVF workers

[mrcnski]

ISSUE

Overview

Original comment by @eskimor:

ru_maxrss ... could be interesting. By checking the amount of memory used, we might get a better heuristic on whether something is likely valid, but the machine has issues or likely invalid (excessive RAM usage). Don't have a clear strategy, seems very brittle at first thought.

We decided that just logging it for now would be nice.

Note: The cpu_time crate also has support for Windows. Not sure if Windows has something like ru_maxrss. Do we care?

Related Links

Follow-up from #6282.

[mrcnski]

Don't think either Windows nor MacOS supports this; see here. We might be able to get similar data with tikv-jemalloc-ctl, as described in paritytech/polkadot-sdk#745.

[bkchr]

We don't really support macos nor windows officially for validators. We merge fixes and ensure that it compiles for these targets, but if it not supporting all features it should be okay.

[vstakhov]

struct rusage is not really defined by POSIX standard, so Darwin and Windows are free to do whatever they like here, I'm afraid. However, at least in the recent OSX versions, struct rusage returned from getrusage call has the field:

 long ru_maxrss;          /* max resident set size */

[mrcnski]

Don't think either Windows nor MacOS supports this; see here.

Actually, I tried it and it seems like MacOS does support max_rss (though not Windows). I'm not really sure what I was looking at in that link TBH. 😅