both nohang tests failed on Ubuntu 18.04 #98

lvitya · 2020-05-28T14:39:11Z

I have installed nohang as described. Then

$ sudo systemctl enable nohang-desktop
$ sudo systemctl start nohang-desktop
$ nohang --memload

And guess what - system freezed. After waiting for 10 minutes I performed a power cycle.
After booting I checked if nohang is running with
$ systemctl list-units

This time I tried
$ tail /dev/zero
The system freezed again.

Is this me doing something wrong or the app? For me as a user this behaviour is not expected.

The text was updated successfully, but these errors were encountered:

hakavlad · 2020-05-28T15:01:47Z

Hi! I'd like to see the output:

$ uname -a

$ cat /proc/swaps

$ cat /proc/sys/vm/swappiness

$ cat /proc/pressure/memory

See also #85

hakavlad · 2020-05-28T15:03:02Z

And i'd like to see the journal since nohang starts.

sudo journalctl -eu nohang-desktop

lvitya · 2020-05-28T16:53:26Z

$ uname -a
Linux viktor-desktop 4.16.3-041603-generic #201804190730 SMP Thu Apr 19 07:32:02 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

$ cat /proc/swaps
Filename				Type		Size	Used	Priority
/dev/dm-0                               partition	8252924	3009020	-2

$ cat /proc/sys/vm/swappiness
60

$ cat /proc/pressure/memory
cat: /proc/pressure/memory: No such file or directory

lvitya · 2020-05-28T16:55:51Z

-- Logs begin at Thu 2020-04-23 14:41:57 EEST, end at Thu 2020-05-28 19:53:05 EEST. --
тра 28 16:51:56 viktor-desktop systemd[1]: Started Sophisticated low memory handler.
тра 28 16:51:56 viktor-desktop nohang-desktop[6284]: config: /etc/nohang/nohang-desktop.conf
тра 28 16:51:56 viktor-desktop nohang-desktop[6284]: WARNING: PSI metrics are not provided by the kernel: [Errno 2] No such file or directory: '/proc/pressure/memory'
тра 28 16:51:56 viktor-desktop nohang-desktop[6284]: Monitoring has started!
-- Reboot --
тра 28 17:05:34 viktor-desktop systemd[1]: Started Sophisticated low memory handler.
тра 28 17:05:36 viktor-desktop nohang-desktop[1243]: config: /etc/nohang/nohang-desktop.conf
тра 28 17:05:36 viktor-desktop nohang-desktop[1243]: WARNING: PSI metrics are not provided by the kernel: [Errno 2] No such file or directory: '/proc/pressure/memory'
тра 28 17:05:36 viktor-desktop nohang-desktop[1243]: Monitoring has started!
-- Reboot --
тра 28 17:21:58 viktor-desktop systemd[1]: Started Sophisticated low memory handler.
тра 28 17:21:59 viktor-desktop nohang-desktop[1270]: config: /etc/nohang/nohang-desktop.conf
тра 28 17:21:59 viktor-desktop nohang-desktop[1270]: WARNING: PSI metrics are not provided by the kernel: [Errno 2] No such file or directory: '/proc/pressure/memory'
тра 28 17:21:59 viktor-desktop nohang-desktop[1270]: Monitoring has started!
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]: Warning threshold exceeded
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]: >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]: Memory status that requires corrective actions:
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]:   MemAvailable [2 MiB, 0.0 %] <= soft_threshold_min_mem [392 MiB, 5.0 %]
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]:   SwapFree [788 MiB, 9.8 %] <= soft_threshold_min_swap [806 MiB, 10.0 %]
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]: Found 26 tasks with non-zero oom_score (except init and self) in 4957ms
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]: TOP-15 tasks by badness:
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]:   Name                PID badness
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]:   --------------- ------- -------
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]:   Web Content       26404     234
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]:   Web Content       26552     226
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]:   Web Content       26160     225

lvitya · 2020-05-28T16:59:37Z

2 reboots in the journal are power cycles after freezes.
After 2nd reboot, I just use PC normally. There are more journal entries after this part.

hakavlad · 2020-05-28T17:28:43Z

PSI required to prevent freezing under heavy swapping.

You could install the new kernel (4.20+), maybe 5.0 or 5.3.

Demo: https://youtu.be/Y6GJqFE_ke4

hakavlad · 2020-05-28T17:30:09Z

https://www.kernel.org/doc/html/latest/accounting/psi.html

hakavlad · 2020-05-28T17:40:32Z

Also disabling swap space can help you to prevent freezeng: this will help to avoid prolonged swapping. Demo without swapping: https://youtu.be/UCwZS5uNLu0

lvitya · 2020-05-28T18:03:14Z

You could install the new kernel (4.20+), maybe 5.0 or 5.3.

It is ok for me. However, according to README PSI is not required.

Also disabling swap space can help you to prevent freezing: this will help to avoid prolonged swapping.

What if I have only 8 GB of RAM? I remember I read in some article that it is better to have swap enabled. I only have SSD in this system. Theoretically, swap speed should be fast. Could you please clarify about swap?

hakavlad · 2020-05-28T18:46:32Z

README PSI is not required

for basic usage, i e to handle low MemAvailable/SwapFree.

Theoretically, swap speed should be fast

As you see, it is not fast enough to prevent freezing.

Could you please clarify about swap?

Yes, later, plz wait.

hakavlad · 2020-05-28T18:48:01Z

I read in some article that it is better to have swap enabled

Yes, the article is https://chrisdown.name/2018/01/02/in-defence-of-swap.html

hakavlad · 2020-05-30T19:02:46Z

Is this me doing something wrong or the app?

App works as intended. It responds to low MemAvailable and SwapFree. As you see:

тра 28 17:58:39 viktor-desktop nohang-desktop[1270]: Memory status that requires corrective actions:
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]:   MemAvailable [2 MiB, 0.0 %] <= soft_threshold_min_mem [392 MiB, 5.0 %]
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]:   SwapFree [788 MiB, 9.8 %] <= soft_threshold_min_swap [806 MiB, 10.0 %]
тра 28 17:58:39 viktor-desktop nohang-desktop[1270]: Found 26 tasks with non-zero oom_score (except init and self) in 4957ms

My fault is that the documentation is terrible and I did not explain what "basic usage" is.

After waiting for 10 minutes I performed a power cycle

I think that the problem would be resolved if you waited 30 minutes. The memory will be freed after filling the swap and killing the victim.

The situation can be improved by reducing the size of the swap space. Small swap space may be filled faster and corrective action will happen faster.

Using PSI metrics allows you to detect freezing faster and perform corrective action less than one minute.

lvitya · 2020-05-30T19:39:11Z

Thank you for the detailed replies.
I'm looking for a good way to install a kernel with PSI support. Previously I used UKUU but recently the app goes to "paid only". I can't install the app from the usual PPA. I want to find some repeatable way to install a kernel. The CLI way is maybe even better.

I plan to post the update here after I have some progress. This may take some time depending on how much time I can dedicate to this issue.

lvitya · 2020-05-30T19:46:14Z

I think that the problem would be resolved if you waited 30 minutes. The memory will be freed after filling the swap and killing the victim.

Actually I thought one of the features of earlyoom, nohang and alike is avoiding waiting such a long time.
I was not aware that they require PSI support to act fast.

hakavlad · 2020-05-30T19:55:48Z

I'm looking for a good way to install a kernel with PSI support.

See https://itsfoss.com/ubuntu-hwe-kernel/

Maybe you would run sudo apt install --install-recommends linux-generic-hwe-18.04 xserver-xorg-hwe-18.04 to install 5.3 with PSI support.

Hardware Enablement Stacks (HWE) are incorporated into installers for select Ubuntu LTS (Long Term Support) point releases. It is a special Ubuntu feature that provides an LTS release with hardware support introduced in newer Ubuntu releases.

https://packages.ubuntu.com/en/bionic/linux-generic-hwe-18.04

hakavlad · 2020-05-30T20:00:03Z

Actually I thought one of the features of earlyoom, nohang and alike is avoiding waiting such a long time.

Without swap space you can avoid waiting. nohang without PSI works like earlyoom: it responds to MemAvailable and SwapFree.
If you want to avoid long-time freezing with swap space you should have PSI support.

lvitya · 2020-06-03T17:45:21Z

Maybe you would run sudo apt install --install-recommends linux-generic-hwe-18.04 xserver-xorg-hwe-18.04 to install 5.3 with PSI support.

So I have installed Linux kernel 5.3.0 according to provided suggestion.
Now nohang tests don't lead to Ubuntu freeze. Thank you.

One more question is why after killing the hog process the memory isn't freed?
See, for example, this log after tail test.
https://gist.github.com/lvitya/7241dda5b3f7723d84cfec41c55779f6

Memory status after implementing a corrective action:
MemAvailable: 164.9 MiB, SwapFree: 1467.9 MiB

hakavlad · 2020-06-13T08:16:48Z

One more question is why after killing the hog process the memory isn't freed?

Processes do not free memory immediately after receiving a signal, even if it is SIGKILL. Freeing up memory can take up to several seconds. See also rfjakob/earlyoom#128 (comment)

alexmyczko · 2023-03-31T15:03:39Z

Do you still run Ubuntu 18.04? Or did you meanwhile upgrade to 20.04 or 22.04?

lvitya · 2023-03-31T15:35:44Z

I upgraded to 20.04.

alexmyczko · 2023-03-31T15:47:19Z

@lvitya what is holding you from going 22.04?

lvitya · 2023-03-31T16:21:43Z

The transition introduces a bunch of problems. The conclusion is based on my previous experience.
Too many things change at once. I still miss some Unity features.
Something got broken. For example, manually added repositories don't work anymore.
I should be ready to spend an uncertain amount of time fixing my user and dev environments.

alexmyczko · 2023-03-31T16:56:16Z

I agree something is fishy going on with Ubuntu, here is my solutions. https://github.com/alexmyczko/autoexec.bat/tree/master/config.sys

kecsap mentioned this issue Jun 5, 2020

Unable to get warning notifications #100

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

both nohang tests failed on Ubuntu 18.04 #98

both nohang tests failed on Ubuntu 18.04 #98

lvitya commented May 28, 2020

hakavlad commented May 28, 2020

hakavlad commented May 28, 2020

lvitya commented May 28, 2020

lvitya commented May 28, 2020

lvitya commented May 28, 2020

hakavlad commented May 28, 2020

hakavlad commented May 28, 2020

hakavlad commented May 28, 2020

lvitya commented May 28, 2020 •

edited

Loading

hakavlad commented May 28, 2020 •

edited

Loading

hakavlad commented May 28, 2020

hakavlad commented May 30, 2020

lvitya commented May 30, 2020

lvitya commented May 30, 2020

hakavlad commented May 30, 2020 •

edited

Loading

hakavlad commented May 30, 2020

lvitya commented Jun 3, 2020

hakavlad commented Jun 13, 2020

alexmyczko commented Mar 31, 2023

lvitya commented Mar 31, 2023

alexmyczko commented Mar 31, 2023

lvitya commented Mar 31, 2023 •

edited

Loading

alexmyczko commented Mar 31, 2023

both nohang tests failed on Ubuntu 18.04 #98

both nohang tests failed on Ubuntu 18.04 #98

Comments

lvitya commented May 28, 2020

hakavlad commented May 28, 2020

hakavlad commented May 28, 2020

lvitya commented May 28, 2020

lvitya commented May 28, 2020

lvitya commented May 28, 2020

hakavlad commented May 28, 2020

hakavlad commented May 28, 2020

hakavlad commented May 28, 2020

lvitya commented May 28, 2020 • edited Loading

hakavlad commented May 28, 2020 • edited Loading

hakavlad commented May 28, 2020

hakavlad commented May 30, 2020

lvitya commented May 30, 2020

lvitya commented May 30, 2020

hakavlad commented May 30, 2020 • edited Loading

hakavlad commented May 30, 2020

lvitya commented Jun 3, 2020

hakavlad commented Jun 13, 2020

alexmyczko commented Mar 31, 2023

lvitya commented Mar 31, 2023

alexmyczko commented Mar 31, 2023

lvitya commented Mar 31, 2023 • edited Loading

alexmyczko commented Mar 31, 2023

lvitya commented May 28, 2020 •

edited

Loading

hakavlad commented May 28, 2020 •

edited

Loading

hakavlad commented May 30, 2020 •

edited

Loading

lvitya commented Mar 31, 2023 •

edited

Loading