Regular load spikes

[aueelis]

A while ago, I noticed regular load spikes on my machine. Investigating this issue further, I registered spikes of ~5.5 every 125 minutes. After reducing the size of my script, the spikes reduced to ~4 every 125 minutes. After killing conky, the spikes disappear.

conkyrc.txt

[plikhari]

There are many things happening when you are running a program. 125 minutes is just over 2 hours - how do you clock this ??? Then again update interval of 1 second - what would happen if you increased it to a higher value. Accuracy and Precision are 2 different things. We normally need to find a balance. Are those load spikes induced due to the update interval ??? Perhaps if you increased that update interval to 3 or 5 - what results do you get??

[aueelis]

I'm monitoring the load on my system with watch by logging the contents of /proc/loadavg:

watch -p -n 1 'echo -n $(date +%s) " " | tee -a loadlog && cat /proc/loadavg | tee -a loadlog'

At the moment I'm trying to investigate, which conky widget contributes to this spike. Running only the first part of the script down to the cpugraph (i.e. ll. 1-76) doesn't show a measurable spike (however, gdm produces an unrelated background load of 0.3 on this machine...). The measurement with line 77 (loadgraph) is running now.

Later, I'll play with the update interval.

[aueelis]

I couldn't find a specific command, which could cause these spikes.

Playing with the update interval (0.5s, 1s and 5s) didn't change the distance between the spikes.

[aueelis]

I think, it is related to the execi command!

The following setup produced no measurable spikes:

${font $template0}SYSTEM ${hr 2}
${font $template1}$sysname $kernel $alignr $machine
Host:$alignr$nodename
Uptime:$alignr$uptime

${font $template0}PROZESSOR ${hr 2}
${font $template1}${cpugraph cpu0}
${voffset -6}${loadgraph 13,200 FFFFFF FFFFFF 8}
Last:$alignr${loadavg 1} / ${loadavg 2} / ${loadavg 3}
CPU1: ${alignr 15}${cpu cpu1}%$alignr${cpubar cpu1 6,110}
CPU2: ${alignr 15}${cpu cpu2}%$alignr${cpubar cpu2 6,110}
CPU3: ${alignr 15}${cpu cpu3}%$alignr${cpubar cpu3 6,110}
CPU4: ${alignr 15}${cpu cpu4}%$alignr${cpubar cpu4 6,110}

This setup produced spikes:

Updates: ${execi 300 cat "/var/log/pending_updates.log" | wc -l}$alignr Stand: ${execi 300 date -r /var/log/pending_updates.log +%R}
CPU1: ${execi 5 sensors coretemp-isa-0000 | grep "Core 0" | cut -b16-17}°C
CPU2: ${execi 5 sensors coretemp-isa-0000 | grep "Core 1" | cut -b16-17}°C
CPU3: ${execi 5 sensors coretemp-isa-0000 | grep "Core 2" | cut -b16-17}°C
CPU4: ${execi 5 sensors coretemp-isa-0000 | grep "Core 3" | cut -b16-17}°C
Lüfter: $alignr${execi 5 sensors f71858fg-isa-0a00 | grep "fan1:" | cut -b14-17} / ${execi 5 sensors f71858fg-isa-0a00 | grep "fan2:" | cut -b14-17}

Unfortunately, I'm absolutely not sure, where to go from here... Any suggestions?

[aueelis]

Closing for now: After the last update of gdm, the background load was reduced to around 0.06. This enabled some deeper analysis (e.g. deconvolution and FFT) with and without conky. Even without conky, I notice regular spikes every 25 minutes (not 125).

My conclusion (for now) is that conky is not the cause, but an amplifier of the load spikes.