Re: Too many pythons in poudriere

From: Mark Millard via freebsd-ports <freebsd-ports_at_freebsd.org>
Date: Thu, 08 Jul 2021 22:47:17 UTC
From: Tatsuki Makino <tatsuki_makino_at_hotmail.com> wrote on
Date: Fri, 9 Jul 2021 05:59:52 +0900 :

> bob prohaska wrote on 2021/07/09 02:54:
> > Started a new swaplog:
> 
> Connect a USB mass storage device that you have no use for.
> Mount the file system in it.
> Create a file in that file system like dd if=/dev/zero of=/mnt/somefile.
> Attach it to the memory disk like mdconfig /mnt/somefile.
> Add it to the swap.

Bob already has the swap space:

Device          1K-blocks     Used    Avail Capacity
/dev/da0s2b       1843200    26944  1816256     1%
/dev/mmcsd0s2b    1843200    26956  1816244     1%
Total             3686400    53900  3632500     1%

This is nearly as much as is allowed without getting
swapon-time-frame notices of the form:

warning: total configured swap (??? pages) exceeds maximum recommended amount (??? pages).

Bob may be stuck with a choice among:

A) running the RPi3B (1 GiByte RAM) mistuned such that
   the above messages are reported (or adjusting other
   kernel tradeoffs explicitly to avoid the message)
vs.
B) building www/chromium on another machine that has
   more RAM then installing the package built on the
   RPi3B for later use
vs.
C) giving up on building www/chromium at all, using a
   pre-built package
vs.
D) giving up on even using www/chromium on the RPi3B.

Bob is familiar with doing (A). I'm the one that
requested that he try not doing so as part of trying
to isolate a problem building devel/llvm10 . That
problem was later identified to be a different issue.
(A) does make other kernel resource tradeoffs,
either implicitly or explicitly. I do not know how
to well manage the tradeoffs involved so I never
recommend (A). Others may know what they are doing
for making such tradeoffs.

However, even if Bob does (A) there may be other issues
that he will run into. Form build attempt to build
attempt he is having problems leading to massive swap
space usage at differing points in the build. The same
step sometimes works fine and other times fails. This
suggests another issue may be present (racy leak?).

> Ask Mark to explain the details to you :)


I would never recommend using files for swap space on
FreeBSD, only swap partitions. See comments 7 and 8 of:

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=206048

for notes from Konstantin Belousov that lead me to
not do so. (8 is a reference to other material.)


===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)