You are not logged in.
Hi, I currently often experience the following scenario:
During working on my laptop, the UI suddenly freezes and the LED indicating disk activity turns on and stays on.
The system is not completely frozen, but everything updates really slowly, maybe every few seconds. Since this happens about once or twice a day, I currently leave iotop running, and I see that almost ALL processes suddenly start to read huge ammounts of data from disk, even simple processes like xfce4-terminal, i3bar and nm-applet. Below you can see a screenshot taken shortly after the disk io ceased.
The screenshot shows the output of "iotop -oPa" on the left and the output of htop on the right. As iotop indicates, many processes had read multiple gigabytes from disk which caused the laptop to lag/freeze for 20 minutes. Also the load still appears very high. I use a fast SSD but still the disk IO caused the system to freeze almost completely.
I don't know what causes this behavior and am grateful for any tips which could help to fix this issue. I suspect it has something to do with me not having a swap file/partition and the system running out of memory, but I'm not sure. htop did not show a full memory in the screenshot, but maybe some processes were already killed due to insufficient memory at that time.
I don't think this should be a normal behavior if running out of memory, so I suspect I misconfigured something.
I use Linux version 4.3.2-1-ck on x86_64 Intel i5-3317U, my memory is 8GB, of which 4 GB are mounted as a tmpfs. Here are the relevant lines from /etc/fstab:
tmpfs /tmp tmpfs nodev,nosuid,size=4G 0 0
tmpfs /var/log tmpfs nodev,nosuid,size=50M 0 0
df -h outputs:
dev 3,8G 0 3,8G 0% /dev
run 3,8G 1,1M 3,8G 1% /run
/dev/sda6 40G 30G 7,9G 79% /
tmpfs 3,8G 150M 3,7G 4% /dev/shm
tmpfs 3,8G 0 3,8G 0% /sys/fs/cgroup
tmpfs 4,0G 707M 3,4G 18% /tmp
tmpfs 50M 36K 50M 1% /var/log
/dev/sda7 77G 70G 2,8G 97% /home
/home/jan/.Private 77G 70G 2,8G 97% /home/jan
tmpfs 770M 60K 770M 1% /run/user/1000
If you need more Information about my setup, just ask.
I really appreciate any help you can provide.
Last edited by nullundnix (2016-01-10 22:38:52)
Offline
Sounds like you run out of memory and the oom killer doesn't trigger (quickly enough). You could see if enabling Magic SysRq and invoking the oom killer manually via Alt+SysRq+f helps.
Offline
Sounds like you run out of memory and the oom killer doesn't trigger (quickly enough). You could see if enabling Magic SysRq and invoking the oom killer manually via Alt+SysRq+f helps.
Thanks for this tip, I will try that, I wasn't aware of Magic SysRq, I think this might help to get the laptop to respond again. Still I would like to know why this behavior occurs in the first place.
Offline
Sounds like you are thrashing
When this occurs, what is the output of free ?
What is the output of cat /proc/swaps ?
Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
Sometimes it is the people no one can imagine anything of who do the things no one can imagine. -- Alan Turing
---
How to Ask Questions the Smart Way
Offline
Sounds like you are thrashing
When this occurs, what is the output of free ?
What is the output of cat /proc/swaps ?
I just provoked this scenario just by opening some more tabs in chromium. The output of your commands was:
% free
total used free shared buff/cache available
Mem: 7877000 5859524 52040 1812308 1965436 95768
Swap: 0 0 0
% cat /proc/swaps
Filename Type Size Used Priority
As I mentioned earlier, I don't currently use a swap file/partition, so /proc/swaps does not show any swaps. Would using swap help in this situation?
Offline
As I mentioned earlier, I don't currently use a swap file/partition, so /proc/swaps does not show any swaps. Would using swap help in this situation?
Missed that More coffee for ewaller.
Yes, I think it would be help to have a swap partition in this case. I don't like running while depending on swap, and my system almost never uses it unless I am doing something extreme -- but it does provide a way to degrade a bit more gracefully.
And yes, your output shows you were low on memory
Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
Sometimes it is the people no one can imagine anything of who do the things no one can imagine. -- Alan Turing
---
How to Ask Questions the Smart Way
Offline
You have 93 MB of memory available. Either an oom_killer kills something or programs fight for the available memory.
Adding swap would surely help (you could also use zswap or zram). Quite a lot of memory pages may be left untouched at all and reside in swap instead of occupying RAM.
Offline
Okay, I just created 2GB swap file and activated it, not it starts swapping when memory runs out and it at least does not cause the system to freeze.
I guess I will also shrink my /tmp tmpfs in order to not run out of memory all that often. Thanks for your help, I think I now understand what the problem was
Offline
Contents of tmpfs can also be swapped out. And tmpfs size only determines how much memory (RAM + swap) can be taken at maximum.
For reference, the output of "free -m" on my desktop right now, with zswap and with everything working fluidly:
total used free shared buff/cache available
Mem: 2019 1382 154 88 482 508
Swap: 3071 864 2207
As one can see, stuff in swap is almost half my RAM.
Last edited by lucke (2016-01-12 17:05:32)
Offline