Memory leak in Lite Widget?

[Richard B]

LL5 is excellent. I am having issues getting my head around EUFI but that is another matter. The following problem was, I am pretty sure, also present in LL4.

After having a problem with a machine "seizing up" if left on for several days I investigated and think there is a problem with the Lite Widget (or conky, which it uses) grabbing more and more memory over time. I have confirmed this using LL5.0 on several machines, with both Intel and AMD processors and from 2GB to 12GB RAM. Some were totally new installs. My tests use the following script:

Code:

#!/bin/bash
# litewsize.sh is to report conky's memory use every hour
# Licence: Any legal use that does not harm anyone is permitted
echo
echo litewsize.sh reports conky\'s memory use every hour
echo
echo runs top -d 3600 -b \|grep conky
echo  . PID  USER .... PR . NI VIRT ... RES .. SHR . S . %CPU %MEM . TIME+ . COMMAND
top -d 3600 -b |grep conky

These are the first results using a machine with 4GB RAM. The script was started after 2 hours "up" time.

Code:

litewsize.sh reports conky's memory use every hour


runs top -d 3600 -b |grep conky
. PID USER .... PR . NI VIRT ... RES .. SHR . S . %CPU %MEM . TIME+ . COMMAND
2096 nomaz     20   0  631176  37620  13776 S   0.0  1.0   1:56.87 conky
2096 nomaz     20   0  639876  46280  13776 S   1.7  1.2   2:56.29 conky
2096 nomaz     20   0  649204  55616  13776 S   1.7  1.4   3:56.68 conky
2096 nomaz     20   0  657784  64200  13776 S   1.8  1.7   5:00.10 conky
2096 nomaz     20   0  666536  72916  13776 S   1.7  1.9   6:02.34 conky
2096 nomaz     20   0  676128  82576  13776 S   1.8  2.1   7:05.89 conky
2096 nomaz     20   0  684736  91204  13776 R   1.9  2.3   8:15.05 conky
2096 nomaz     20   0  693316  99812  13776 S   1.9  2.6   9:22.70 conky
2096 nomaz     20   0  702028 108416  13776 S   1.9  2.8  10:30.71 conky
2096 nomaz     20   0  710864 117244  13776 S   1.9  3.0  11:40.61 conky

Note that memory used by conky has gone from 1% to 3% between 2 and roughly 12 hours. (For whatever reason the timing of the tests is only approximately every 3600 seconds.)


Leaving this running, the RAM usage after 700 hours was nearly 33% as shown here:

Code:

2096 nomaz     20   0 2717092 1.213g   2260 S   1.4 32.8 700:17.74 conky

I think there is a problem. Pursuing it further is beyond me but I would be happy to test any proposed solution.


Regards to Jerry and all developers and users.
Richard Bowers.

[Artim]

Cony is overrated anyway. There's an Xfce panel add-on that can monitor and display CPU and RAM much more efficiently. I replace Conky with the Xfce panel thing on every install.

[Moltke]

LL5 is excellent. I am having issues getting my head around EUFI but that is another matter. The following problem was, I am pretty sure, also present in LL4.

After having a problem with a machine "seizing up" if left on for several days I investigated and think there is a problem with the Lite Widget (or conky, which it uses) grabbing more and more memory over time. I have confirmed this using LL5.0 on several machines, with both Intel and AMD processors and from 2GB to 12GB RAM. Some were totally new installs. My tests use the following script:

Code:

#!/bin/bash
# litewsize.sh is to report conky's memory use every hour
# Licence: Any legal use that does not harm anyone is permitted
echo
echo litewsize.sh reports conky\'s memory use every hour
echo
echo runs top -d 3600 -b \|grep conky
echo  . PID  USER .... PR . NI VIRT ... RES .. SHR . S . %CPU %MEM . TIME+ . COMMAND
top -d 3600 -b |grep conky

These are the first results using a machine with 4GB RAM. The script was started after 2 hours "up" time.

Code:

litewsize.sh reports conky's memory use every hour


runs top -d 3600 -b |grep conky
. PID USER .... PR . NI VIRT ... RES .. SHR . S . %CPU %MEM . TIME+ . COMMAND
2096 nomaz     20   0  631176  37620  13776 S   0.0  1.0   1:56.87 conky
2096 nomaz     20   0  639876  46280  13776 S   1.7  1.2   2:56.29 conky
2096 nomaz     20   0  649204  55616  13776 S   1.7  1.4   3:56.68 conky
2096 nomaz     20   0  657784  64200  13776 S   1.8  1.7   5:00.10 conky
2096 nomaz     20   0  666536  72916  13776 S   1.7  1.9   6:02.34 conky
2096 nomaz     20   0  676128  82576  13776 S   1.8  2.1   7:05.89 conky
2096 nomaz     20   0  684736  91204  13776 R   1.9  2.3   8:15.05 conky
2096 nomaz     20   0  693316  99812  13776 S   1.9  2.6   9:22.70 conky
2096 nomaz     20   0  702028 108416  13776 S   1.9  2.8  10:30.71 conky
2096 nomaz     20   0  710864 117244  13776 S   1.9  3.0  11:40.61 conky

Note that memory used by conky has gone from 1% to 3% between 2 and roughly 12 hours. (For whatever reason the timing of the tests is only approximately every 3600 seconds.)


Leaving this running, the RAM usage after 700 hours was nearly 33% as shown here:

Code:

2096 nomaz     20   0 2717092 1.213g   2260 S   1.4 32.8 700:17.74 conky

I think there is a problem. Pursuing it further is beyond me but I would be happy to test any proposed solution.


Regards to Jerry and all developers and users.
Richard Bowers.

What do you mean by "having a problem with a machine "seizing up"? Does your system/machine get short of RAM? What's swappiness value? 10? 60? I wonder, is this really accurate? I mean, don't get me wrong but how can you tell it is? How does the math for those numbers work? Is it cumulative? I see that at the start of the test conky RAM usage is 1.0 - here I get 0.0 - and after 11 hours is 3.0, how is that program/script making those numbers? And then after 700 hours is 32.8? That can't be right. I think, one should first make sure those numbers are accurate, I mean, that the script/program is not making a grand total by adding all the values up per hour, which is not useful for the purpose. So maybe, launching a terminal or a script with

Code:

watch -n 3600 free -h

might help to keep track of RAM usage and compare results. I might be wrong though. FWIW, I'm just trying to understand how the whole process in this test works; how the math for those numbers works.  Nothing else. 

[Valtam]

I'd be more concerned if this was deployed on a server. Given our target audience will only use their pc for a few hours at a time it seems unurgent. Happy for folks to dig deeper and pin point it.

Sent from my mobile phone using Tapatalk