Thread: [Monitorix-general] lmsensors configuration
Monitorix is a system monitoring tool
Brought to you by:
mikaku
Menu
▾
▴
monitorix-general
|
From: Steve V. <nu...@ho...> - 2022-04-21 13:06:15
Attachments:
image001.png
|
Hello, I am new to monitorix, having learned of and installed it only yesterday, so please forgive my ignorance.
I have managed to get most of the graphs configured the way I want them and configured the server to listen on only an internal network address. I have had some trouble with the LMSENSORS graph for two reasons: My system doesn't have any of the sensors required for some of the graphs and uses the same name for multiple sensor values. Here's the output from a sensors command on my machine:
steve@mail:/etc/monitorix$ sensors
dell_smm-virtual-0
Adapter: Virtual device
Other Fan: 3114 RPM
Processor Fan: 3148 RPM
Processor Fan: 3137 RPM
Ambient: +25.0°C
Ambient: +24.0°C
Ambient: +27.0°C
Ambient: +24.0°C
radeon-pci-0200
Adapter: PCI adapter
temp1: +56.0°C
coretemp-isa-0000
Adapter: ISA adapter
Core 0: +54.0°C (high = +87.0°C, crit = +103.0°C)
Core 1: +55.0°C (high = +87.0°C, crit = +103.0°C)
Core 2: +55.0°C (high = +87.0°C, crit = +103.0°C)
Core 3: +56.0°C (high = +87.0°C, crit = +103.0°C)
And here's what the graph looks like:
[cid:image001.png@01D8555D.FED04090]
And finally, here's what I've got for configuration:
# LMSENS graph
# -----------------------------------------------------------------------------
<lmsens>
<list>
core0 = Core 0
core1 = Core 1
core2 = Core 2
core3 = Core 3
mb0 = Ambient
#cpu0 = CPU Temp
fan0 = Other Fan
fan1 = Processor Fan
#fan2 = Processor Fan
#volt0 = VCore 1
#volt1 = VCore 2
#volt2 = \+3.3V
#volt3 = \+5V
#volt4 = \+12V
#volt5 = \-12V
#volt6 = \-5V
#volt7 = Battery
gpu0 = temp1
</list>
<desc>
mb0 = Case Ambient
fan0 = Case Fan
fan1 = CPU Fan
gpu0 = Radeon GPU
</desc>
<alerts>
</alerts>
</lmsens>
I'm hoping there is a way to do a few things:
1. Sensibly map all (I hope) of my sensors output to the graphs
2. Find the setting for the min/max values for the various graphs (e.g. the core temps only need 30-70 or so)
3. Remove graphs with no data points from the display.
Hopefully someone here will be able and willing to help! (smile) Thanks in advance.
|
|
From: Jordi S. <jo...@fi...> - 2022-04-21 15:08:58
|
Hello Steve, On 4/21/22 15:05, Steve Valliere wrote: > I’m hoping there is a way to do a few things: > > 1. Sensibly map all (I hope) of my sensors output to the graphs This is out of scope of Monitorix, you need to modify lm_sensors to output different names. Please read this FAQ <https://www.monitorix.org/faq.html#Q125>, which will guide you to <https://github.com/mikaku/Monitorix/issues/131>. > 2. Find the setting for the min/max values for the various graphs (e.g. > the core temps only need 30-70 or so) This would need to introduce the option 'use_nan_for_missing_data' like some other graphs already have (<https://github.com/mikaku/Monitorix/issues/401>). Please, open a new issue in GitHub to help to track this change. > 3. Remove graphs with no data points from the display. That's not possible as the majority of graph layouts in Monitorix are fixed. Regards. -- Jordi Sanfeliu FIBRANET Network Services Provider https://www.fibranet.cat |
|
From: Steve V. <nu...@ho...> - 2022-04-21 18:00:35
|
Thanks, IPMI sounded great, but it seems that my system (a Dell Precision WorkStation R5500, circa 2013) does not support it (Dell appears to have an IPMI driver, but not for linux and/or not for my device.) All of the IPMI detection methods suggested by the freeipmi documentation fail and the ipmi-sensors program runs for a couple minutes and then fails with an internal error (no more detail than that.) Oh well. Using the other suggestion for renaming the lmsensors in the 'sensors' output, I've been able to get all four case temps into the MP and CPU temps graph, and I added my radeon GPU temp to the Core Temperatures graph as well. So voltages and GPU Temps are empty, oh well. Thanks for the help! -----Original Message----- From: Jordi Sanfeliu via Monitorix-general <mon...@li...> Sent: Thursday, April 21, 2022 10:43 AM To: mon...@li... Cc: Jordi Sanfeliu <jo...@fi...> Subject: Re: [Monitorix-general] lmsensors configuration Hello Steve, On 4/21/22 15:05, Steve Valliere wrote: > I’m hoping there is a way to do a few things: > > 1. Sensibly map all (I hope) of my sensors output to the graphs This is out of scope of Monitorix, you need to modify lm_sensors to output different names. Please read this FAQ <https://www.monitorix.org/faq.html#Q125>, which will guide you to <https://github.com/mikaku/Monitorix/issues/131>. > 2. Find the setting for the min/max values for the various graphs (e.g. > the core temps only need 30-70 or so) This would need to introduce the option 'use_nan_for_missing_data' like some other graphs already have (<https://github.com/mikaku/Monitorix/issues/401>). Please, open a new issue in GitHub to help to track this change. > 3. Remove graphs with no data points from the display. That's not possible as the majority of graph layouts in Monitorix are fixed. Regards. -- Jordi Sanfeliu FIBRANET Network Services Provider https://www.fibranet.cat _______________________________________________ Monitorix-general mailing list Mon...@li... https://lists.sourceforge.net/lists/listinfo/monitorix-general |
|
From: mr2dave <mr...@gm...> - 2022-04-21 14:31:47
Attachments:
image001.png
|
Welcome Steve. I don't know what hardware you've got, but I've had better luck with ipmi than lmsens on hardware with a BMC. Try installing a freeipmi package and run "sudo ipmi-sensors" to see if it detects anything. -Dave On Thu, 2022-04-21 at 13:05 +0000, Steve Valliere wrote: > Hello, I am new to monitorix, having learned of and installed it only > yesterday, so please forgive my ignorance. > > I have managed to get most of the graphs configured the way I want > them and configured the server to listen on only aninternal network > address. I have had some trouble with the LMSENSORS graph for two > reasons: My system doesn’t have any of the sensors required for some > of the graphs and uses the same name for multiple sensor values. > Here’s the output from a sensors command on my machine: > > steve@mail:/etc/monitorix$ sensors > dell_smm-virtual-0 > Adapter: Virtual device > Other Fan: 3114 RPM > Processor Fan: 3148 RPM > Processor Fan: 3137 RPM > Ambient: +25.0°C > Ambient: +24.0°C > Ambient: +27.0°C > Ambient: +24.0°C > > radeon-pci-0200 > Adapter: PCI adapter > temp1: +56.0°C > > coretemp-isa-0000 > Adapter: ISA adapter > Core 0: +54.0°C (high = +87.0°C, crit = +103.0°C) > Core 1: +55.0°C (high = +87.0°C, crit = +103.0°C) > Core 2: +55.0°C (high = +87.0°C, crit = +103.0°C) > Core 3: +56.0°C (high = +87.0°C, crit = +103.0°C) > > And here’s what the graph looks like: > > > And finally, here’s what I’ve got for configuration: > > # LMSENS graph > # ------------------------------------------------------------------- > ---------- > <lmsens> > <list> > core0 = Core 0 > core1 = Core 1 > core2 = Core 2 > core3 = Core 3 > mb0 = Ambient > #cpu0 = CPU Temp > fan0 = Other Fan > fan1 = Processor Fan > #fan2 = Processor Fan > #volt0 = VCore 1 > #volt1 = VCore 2 > #volt2 = \+3.3V > #volt3 = \+5V > #volt4 = \+12V > #volt5 = \-12V > #volt6 = \-5V > #volt7 = Battery > gpu0 = temp1 > </list> > <desc> > mb0 = Case Ambient > fan0 = Case Fan > fan1 = CPU Fan > gpu0 = Radeon GPU > </desc> > <alerts> > </alerts> > </lmsens> > > I’m hoping there is a way to do a few things: > 1. Sensibly map all (I hope) of my sensors output to the graphs > 2. Find the setting for the min/max values for the various graphs > (e.g. the core temps only need 30-70 or so) > 3. Remove graphs with no data points from the display. > > Hopefully someone here will be able and willing to help! (smile) > Thanks in advance. > _______________________________________________ > Monitorix-general mailing list > Mon...@li... > https://lists.sourceforge.net/lists/listinfo/monitorix-general |
|
From: Steve V. <nu...@ho...> - 2022-04-21 18:08:28
|
Re #2 (setting Y-axis ranges)... I do not understand the need to add the use nan setting, since to set a sensible range for the bulk of the values on my graph. For example, the core (and GPU) temps on my system are just about always 55C +/- 2C, so why am I required to have a graph whose scale goes from 0 to 60? I was hoping it might auto-adjust the low limit on the scale once the graph filled, but it did not. It looks like both the min and max values for the Y axis are settable parameters for the graphics package, so I'm requesting that we be allowed to set them without hacking into the perl script on our own. But I think I can figure that out if you don't think there's any value to graphs that aren't generally a flat line, um, sorry, I mean if you don't see the value in this option. Thanks, SteveV. -----Original Message----- From: Jordi Sanfeliu via Monitorix-general <mon...@li...> Sent: Thursday, April 21, 2022 10:43 AM To: mon...@li... Cc: Jordi Sanfeliu <jo...@fi...> Subject: Re: [Monitorix-general] lmsensors configuration Hello Steve, On 4/21/22 15:05, Steve Valliere wrote: > I’m hoping there is a way to do a few things: > > 1. Sensibly map all (I hope) of my sensors output to the graphs This is out of scope of Monitorix, you need to modify lm_sensors to output different names. Please read this FAQ <https://www.monitorix.org/faq.html#Q125>, which will guide you to <https://github.com/mikaku/Monitorix/issues/131>. > 2. Find the setting for the min/max values for the various graphs (e.g. > the core temps only need 30-70 or so) This would need to introduce the option 'use_nan_for_missing_data' like some other graphs already have (<https://github.com/mikaku/Monitorix/issues/401>). Please, open a new issue in GitHub to help to track this change. > 3. Remove graphs with no data points from the display. That's not possible as the majority of graph layouts in Monitorix are fixed. Regards. -- Jordi Sanfeliu FIBRANET Network Services Provider https://www.fibranet.cat _______________________________________________ Monitorix-general mailing list Mon...@li... https://lists.sourceforge.net/lists/listinfo/monitorix-general |
|
From: Jordi S. <jo...@fi...> - 2022-04-22 07:54:14
|
Hello, On 4/21/22 20:08, Steve Valliere wrote: > Re #2 (setting Y-axis ranges)... I do not understand the need to add the use nan setting, since to set a sensible range for the bulk of the values on my graph. For example, the core (and GPU) temps on my system are just about always 55C +/- 2C, so why am I required to have a graph whose scale goes from 0 to 60? I was hoping it might auto-adjust the low limit on the scale once the graph filled, but it did not. > It looks like both the min and max values for the Y axis are settable parameters for the graphics package, so I'm requesting that we be allowed to set them without hacking into the perl script on our own. But I think I can figure that out if you don't think there's any value to graphs that aren't generally a flat line, um, sorry, I mean if you don't see the value in this option. > The graphs do autoscaling automatically, this is an RRDtool feature. The problem is that some graphs plot a zero value when there is not value to plot and that forces the graph to scale from 0 to the maximum value represented. This behavior is accepted in the majority of cases. That's the reason why not all the graphs include the option 'use_nan_for_missing_data' which forces to use the value NaN when there is no value to represent. Regards. -- Jordi Sanfeliu FIBRANET Network Services Provider https://www.fibranet.cat |
|
From: Steve V. <nu...@ho...> - 2022-04-22 12:42:29
Attachments:
CoreTemps.png
|
Sorry for being a pest, but I'm trying to figure out what I've done wrong. After this I promise to shut up and go away. (smile) I enabled the hourly graphs and got this for the last hour's CPU Core temperatures (image attached). When I've seen a missing/zero value, there has always been a (near) vertical line drawn between the X axis and the 'actual' values. The attached graph doesn't show any of that and yet did not auto scale. If you look closely, you can see that all five tracks are plotted in the very narrow 'track blob' at the top. So, how did I defeat the auto-scaling feature? Or is the graph /using/ a zero value (that it invented, since 'sensors' never outputs 0C for a CPU Core temp!) and electing not to draw the associated line segment(s)? [Side note: I tried switching to text mode output to look at the values, but all of the selections I have tried fail with messages like "ERROR: while fetching /var/lib/monitorix/lmsens.rrd: resolution: value must be (suffixed) positive number" on the HOURLY setting -- text mode seems to work fine for daily and above] -----Original Message----- From: Jordi Sanfeliu via Monitorix-general <mon...@li...> Sent: Friday, April 22, 2022 3:54 AM To: mon...@li... Cc: Jordi Sanfeliu <jo...@fi...> Subject: Re: [Monitorix-general] lmsensors configuration Hello, On 4/21/22 20:08, Steve Valliere wrote: > Re #2 (setting Y-axis ranges)... I do not understand the need to add the use nan setting, since to set a sensible range for the bulk of the values on my graph. For example, the core (and GPU) temps on my system are just about always 55C +/- 2C, so why am I required to have a graph whose scale goes from 0 to 60? I was hoping it might auto-adjust the low limit on the scale once the graph filled, but it did not. > It looks like both the min and max values for the Y axis are settable parameters for the graphics package, so I'm requesting that we be allowed to set them without hacking into the perl script on our own. But I think I can figure that out if you don't think there's any value to graphs that aren't generally a flat line, um, sorry, I mean if you don't see the value in this option. > The graphs do autoscaling automatically, this is an RRDtool feature. The problem is that some graphs plot a zero value when there is not value to plot and that forces the graph to scale from 0 to the maximum value represented. This behavior is accepted in the majority of cases. That's the reason why not all the graphs include the option 'use_nan_for_missing_data' which forces to use the value NaN when there is no value to represent. Regards. -- Jordi Sanfeliu FIBRANET Network Services Provider https://www.fibranet.cat _______________________________________________ Monitorix-general mailing list Mon...@li... https://lists.sourceforge.net/lists/listinfo/monitorix-general |
|
From: Jordi S. <jo...@fi...> - 2022-04-23 07:21:04
|
Hello, On 4/22/22 14:42, Steve Valliere wrote: > Sorry for being a pest, but I'm trying to figure out what I've done wrong. After this I promise to shut up and go away. (smile) Steve, you are very welcome :-), this list has a very low activity, almost only used for version announcements. Nowadays, most people seem to prefer GitHub to ask for issues, questions, etc. Anyway, it's fine to see someone writing here. > I enabled the hourly graphs and got this for the last hour's CPU Core temperatures (image attached). When I've seen a missing/zero value, there has always been a (near) vertical line drawn between the X axis and the 'actual' values. The attached graph doesn't show any of that and yet did not auto scale. If you look closely, you can see that all five tracks are plotted in the very narrow 'track blob' at the top. > > So, how did I defeat the auto-scaling feature? Or is the graph /using/ a zero value (that it invented, since 'sensors' never outputs 0C for a CPU Core temp!) and electing not to draw the associated line segment(s)? I'm sorry, now I realized that the option 'use_nan_for_missing_data' was already recently included in the 'lmsens.pm' module. See here <https://github.com/mikaku/Monitorix/pull/406>. So you only need to add the following options in your <lmsens> section: use_nan_for_missing_data = y limit = 100, 100, 100, 100, 100 and wait until all data plotted (in your selected time frame) is changed to use NaN for all zero values. That is, your graph will start autoscaling correctly after an hour in the hourly graph and after a day in the daily graph and so on. Check the following issue <https://github.com/mikaku/Monitorix/pull/403> to read a similar problem a user had but with a different graph. > [Side note: I tried switching to text mode output to look at the values, but all of the selections I have tried fail with messages like "ERROR: while fetching /var/lib/monitorix/lmsens.rrd: resolution: value must be (suffixed) positive number" on the HOURLY setting -- text mode seems to work fine for daily and above] Yes, text version might well support only standard time frames (daily, weekly, etc.), hourly is a 'tricky' time frame. Regards. -- Jordi Sanfeliu FIBRANET Network Services Provider https://www.fibranet.cat |
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X