Cheshire Cat Computing

I just tried this template on a linux server, couple of comments.

Firstly I ran into the H1 issue that henrik did, using his version of the script fixed that.
Secondly "free" memory on a linux box isn't actually a very useful figure since linux will use as much as it can for disk cache. So generally free+cached is more interesting than free.

The issue with indexmaker and configs generated has to do with the PageTop entry not being encapsulated in heading tags in the generated config file.

For example:

PageTop[my-server-cpu]: my-server CPU usage

needs to be:

PageTop[my-server-cpu]: <H1>my-server CPU usage</H1>


I manually went through the generated config file and added the H1 tags to everywhere PageTop was. After that I was able to run indexmaker on the file with no errors.

Great template otherwise!

I have made some more minor updates to the template:

1) The load average is no longer quantized to integers.
2) One more low pass filtered load average is presented as a blue line
3) The max valid load average has been increased from 1000 to 2000.

regards Henrik

Thanks for this; I'll update the main copy. Since we don't use indexmaker I'd not spotted the <H1> issue before, so it's good to get the feedback.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Thanks for providing this.

I would think it might be better to use " / 1000" on the Target[] line to convert to degrees Celsius, rather than using Factor[]:0.001, in the CPU temperature definition, so that the values are stored in Celsius. If you do this, it makes it easier to potentially use the c2fin options to display in Farenheit (for the benefit of Americans). I also think that (while technicaly Kelvin is the SI) it's still best to store the data in Celsius rather man in millidegrees. I think that it's always best to do the calculations on the Target line rather than using Factor if you have the choice.

S

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Sorry about the Factor in my first version. I first hacked the host template file together, then returned to this forum and was reminded about the preference for SI units in the thread about DELL UPS host template.

Included with this message is a fixed host template which no longer relies on Factor to fix the scaling.

Even though I have added some host template files to this forum I must admit that I am no expert in template writing. I would not be able to write a new host template from scratch without heavy consulting of documentation. Instead my method of writing or modifying templates usually are: copy and paste from something similar and then more or less random hacking with trial and error until the template behaves as expected.

regards Henrik

I have now updated the host template to also support fan speeds and voltages. As before, the host template is fully backwards compatible for hosts which does not support lm-sensors.

As the lm-sensors MIB states that the temperature shall be given in millidegrees Celsius and the voltages in milliVolt I have also reverted back to the usage of Factor for temperatures and also for the new voltage functionality.
As far as the temperatures are concerned, I think it wouldn't harm much to use "/1000" on the Target instead, but for some of the voltages those fractions really get significant. You would not want to lose the precision and be unable to tell the difference between 3.31 V and 3.0 V.

I have also now gotten some experience from a little weirdness of net-snmp:

If the lm-sensors gives the same name for two different sensors, like "fan1" on both the motherboard and the graphics card, using the lm-sensors command "sensors" will display the values of both these "fan1". However, snmpwalk and then also of course my host template will only show the value of one of these sensors. The solution to this problem is to rename at least one of these sensors, this is done by editing a file named something like /etc/sensors.conf. Once that file has been edited, snmpd should be restarted on that host. After that snmpwalk will be able to tell the sensors apart and my host template will give graphs for both sensors.

regards Henrik