1 |
Your answer seems to be on the right track I expected. |
2 |
|
3 |
|
4 |
|
5 |
|
6 |
|
7 |
At 2022-09-17 21:08:21, "Rich Freeman" <rich0@g.o> wrote: |
8 |
>On Sat, Sep 17, 2022 at 8:21 AM johnstrass <johnstrass@×××.com> wrote: |
9 |
>> |
10 |
>> |
11 |
>> Why is the logind so fragile? |
12 |
> |
13 |
>Have you checked your logs. I'm guessing that the kernel OOM killer |
14 |
>is killing it, and it is kind of hard for a process to not die when |
15 |
>the kernel kills it. |
16 |
> |
17 |
>> Why cannot it be brought up again after the memeory become available again? |
18 |
> |
19 |
>I suspect it probably could be - probably not a failure mode upstream |
20 |
>has paid much attention to. If the OOM killer is loose on your |
21 |
>system, general breakage is to be expected. It is actually surprising |
22 |
>though that it didn't go after gcc itself. I'd check the logs. |
23 |
> |
24 |
>You probably could tweak the unit setting so that logind is less |
25 |
>likely to get prioritized. Then it might go after something less |
26 |
>essential like sshd or postfix. :) |
27 |
> |
28 |
>Also possible that it isn't logind itself but something else it uses |
29 |
>for IPC. Haven't looked into the gory details of how it works. |
30 |
> |
31 |
>I'm guessing systemd could be coaxed to shut down without it. It |
32 |
>might actually do that on its own if you give it time, but I haven't |
33 |
>checked the settings. |
34 |
|
35 |
> |
36 |
|
37 |
|
38 |
If you have some clue, I would like to know how to coax it. |
39 |
If you would like to have a look at some logs on my machine, I would love to send them to you next time I meet such problems. |
40 |
|
41 |
|
42 |
>-- |
43 |
>Rich |