Re: can't build rust -- out of swap space

On Tue, 3 Mar 2020 12:56+0300, Yuri Pankov wrote:

> On 03.03.2020 12:44, Trond Endrestøl wrote:
> > On Tue, 3 Mar 2020 12:40+0300, Yuri Pankov wrote:
> > 
> > > On 03.03.2020 11:49, Yuri Pankov wrote:
> > > > With recent pkg fallout, I'm trying to build rust myself first time ever
> > > > (as
> > > > far as I can remember), and it's failing running out of swap on the
> > > > following step:
> > > > 
> > > > Building stage0 std artifacts (x86_64-unknown-freebsd ->
> > > > x86_64-unknown-freebsd)
> > > > running:
> > > > "/usr/ports/lang/rust/work/rustc-1.41.1-src/build/x86_64-unknown-freebsd/stage0/bin/cargo"
> > > > "build" "-Zconfig-profile" "--target" "x86_64-unknown-freebsd"
> > > > "-Zbinary-dep-depinfo" "-j" "1" "-v" "--release" "--frozen" "--features"
> > > > "panic-unwind backtrace compiler-builtins-c" "--manifest-path"
> > > > "/usr/ports/lang/rust/work/rustc-1.41.1-src/src/libtest/Cargo.toml"
> > > > "--message-format" "json-render-diagnostics"
> > > > ^C^C^C
> > > > Build completed unsuccessfully in 0:00:55
> > > > 
> > > > Here I pressed ^C as the build actually continues despite several
> > > > rustdoc,
> > > > python, and other processes being killed.
> > > > 
> > > > swap_pager: out of swap space
> > > > swp_pager_getswapspace(20): failed
> > > > swap_pager: out of swap space
> > > > swp_pager_getswapspace(11): failed
> > > > 
> > > > The system has 32G of RAM and 2GB swap partition (as advised by zfs-auto
> > > > installation option), top shows about 28G of memory free at that moment,
> > > > so
> > > > I'm wondering why the swap is being used, and if 2G should be enough to
> > > > build rust.
> > > 
> > > Looks like I got this wrong, adding a file-backed swap space I was
> > > actually
> > > able to run top, and seeing only 100M of memory being "Free", ~20G memory
> > > reported as "Active", and swap usage constantly growing being consumed by
> > > rustdoc process; something is really wrong here.
> > 
> > Run top(1), hit the o key, type in size, and hit enter to have top
> > sort the process list according to their virtual size. The culprit
> > will eventually work its way to the top.
> 
> Yes, it's rustdoc, and I'm seeing the same behavior as you described in your
> other reply.  What's more interesting, having a little swap, processes get
> killed almost immediately, and the build happily continues resulting
> successful rust package.

In my case, I see this behaviour twice during the build. I'll try and 
reduce my swap partition to 2 GiB and see if that makes a difference 
during the build.

-- 
Trond.