On 12/3/23 01:00, jeremy ardley wrote:
On 3/12/23 15:37, Phil Wyett wrote:
Not to regurgitating info here, I will add a link below that will instruct how
to adjust or disable oom-killer in a sensible manner if you wish to experiment
(your choice and being cautious :-)) if it is in fact the oom-killer algorithm
that is the main cause of your issue.
The top output provided earlier seems to show nearly a gigabyte of swap, a tiny
part of it used. You are right, though, that adding swap will not improve
matters for long.
A small amount of swap probably is a good idea to take care of occasional
memory overcommitment. But on an interactive system, swap thrashing that may
happen with a (or several) greedy enough processes will kill performance for
everything that matters, and if one or more of them is leaking memory, adding
swap (or even more memory) will only delay the collapse.
As a reference point Isolated Web Co is an occasional annoyance here on
machines with well over 64G memory. I kill it without mercy when it appears to
be causing swap.
Regards,
Tom Dial
The issue is not so much Isolated Web Co being terminated, but my entire Mate
session being terminated.
I wouldn't have too much problem if OOM-killer hit Firefox. I have done it
myself when things got slow.
However, I can't see any valid reason for the Mate session to be assassinated?
Or at least be inevitable collateral damage.
