On 14/03/2026 22:13, Samuel Thibault wrote:
Michael Kelly, le lun. 09 mars 2026 20:11:45 +0000, a ecrit:
The system is not actually totally stuck. Pageout processing continues at a
low level.
Just to make sure: does that "low level" just matches the slowness of the disk?
It's not disk performance that is limiting in this scenario. Most tasks
are blocked awaiting page in and the only significant recurring event is
kernel pageout attempting to free memory. Each pageout attempt is
however basically a no-op since ext2fs is itself blocked awaiting page
in. There's a minimum 50ms interval between pageout scans so most of
each 50ms the system is idle. Total system activity is therefore so low
you might think the system was totally locked.
[...]
Might it be that boralus is also behaving this way without it being noticed?
That's possible. For some memory-hungry packages I see several GB of
swap getting consumed, and little useful CPU time is spent, until the
build manages to finish, or times out.
I was referring to the behaviour when there is no mach-defpager running.
I'd like to examine the scenarios that you are describing though to see
if it reveals anything interesting. Which package build is a good
example of what you are describing?
Mike.
