Blog posts

Setting Windows 10 web proxy per-user

There are a couple of GUI routes for setting the system web proxy for Windows 10 – the old control panel page (via Network and Internet – Network Options):

And the new settings style:

Note that the new style does not warn you that you may not be allowed to set the proxy – you can change the settings, but if you select another panel and then go back to Proxy your settings will be gone.

ANSI control codes in Jekyll output breaking emails

Also see Publishing websites with Jekyll, Apache and SVN

If you send console output via email (like, say the output of jekyll build as part of a SVN post-commit hook script) if there are ANSI control characters in the string (e.g. colour codes) this can break things. In this case the mail command (Debian 9 default exim) was only sending text up to the first ANSI code, which meant that the jekyll build error messages (which are yellow and red) were missing.

macOS Catalina, zsh and Anaconda

The default terminal shell in macOS Catalina has changed from bash to zsh. Unfortunately it seems the Anaconda installer doesn’t know about this yet. To fix the install copy the section of code added by Anaconda from

~/.bash_profile

to

~/.zshrc

See this post for example.

FDQNs in FlexNet license files

When trying to query FlexNet licenses using lmutil or similar from systems with a different DNS suffix, make sure that the license file server name contains the FDQN for the SERVER line. If not you can find that lmutils complains that the lmgrd process is not running, even though you can run the actual program with the appropriate license.

For example, with a line in the licence file like

SERVER servername 4eca3b4b8326 1055

running

Monitoring GPU temperatures with nvidia-smi and Check MK (OMD)

In the previous post on this subject we used code from Technische Universität Kaiserslautern to monitor our GPUs using OMD checkmk (now checkmk raw). With some new RTX2080s installed this broke, as the nvidia-smi check doesn’t report anything for ECC errors (rather than 0, as previous cards did). The solution was to remove the ECC checking completely.

The new scripts are:

On the client system in /usr/lib/check_mk_agent/local/ (or plugins/)

nvidia-smi

if which nvidia-smi >/dev/null; then
   echo '<<<nvidia_smi>>>'
   nvidia-smi -q -x > /tmp/.check_mk_nvidia_smi
   cards=$(xml_grep --text_only 'nvidia_smi_log/attached_gpus' /tmp/.check_mk_nvidia_smi | tr -d ' ')
   IFS=$'\n' names=($(xml_grep --text_only 'nvidia_smi_log/gpu/product_name' /tmp/.check_mk_nvidia_smi | tr -d ' '))
   IFS=$'\n' fan_speed=($(xml_grep --text_only 'nvidia_smi_log/gpu/fan_speed' /tmp/.check_mk_nvidia_smi | tr -d ' '))
   IFS=$'\n' gpu_utilization=($(xml_grep --text_only 'nvidia_smi_log/gpu/utilization/gpu_util' /tmp/.check_mk_nvidia_smi | tr -d ' '))
   IFS=$'\n' mem_utilization=($(xml_grep --text_only 'nvidia_smi_log/gpu/utilization/memory_util' /tmp/.check_mk_nvidia_smi | tr -d ' '))
   IFS=$'\n' temperature=($(xml_grep --text_only 'nvidia_smi_log/gpu/temperature/gpu_temp' /tmp/.check_mk_nvidia_smi | tr -d ' '))
   IFS=$'\n' power_draw=($(xml_grep --text_only 'nvidia_smi_log/gpu/power_readings/power_draw' /tmp/.check_mk_nvidia_smi | tr -d ' '))
   IFS=$'\n' power_limit=($(xml_grep --text_only 'nvidia_smi_log/gpu/power_readings/power_limit' /tmp/.check_mk_nvidia_smi | tr -d ' '))
 
   for i in $(seq 1 $cards) ; do
       index=$(($i - 1))
       fan_speed[$index]=${fan_speed[$index]/\%/}
       gpu_utilization[$index]=${gpu_utilization[$index]/\%/}
       mem_utilization[$index]=${mem_utilization[$index]/\%/}
       temperature[$index]=${temperature[$index]/C/}
       power_draw[$index]=${power_draw[$index]/W/}
       power_limit[$index]=${power_limit[$index]/W/}
       echo "$index ${names[$index]} ${fan_speed[$index]} ${gpu_utilization[$index]} ${mem_utilization[$index]} ${temperature[$index]} ${power_draw[$index]} ${power_limit[$index]}"
   done
fi

Don’t forget to make it executable! You also need xml_grep installed.