InfluxDB, Graphana, Promox

[Melanxolik]

Hi All.

I am working on integration proxmox for our monitoring which based on graphana, we are using unfluxdb.

Probably does somebody has dashboards? I tried to create several dashboards, but faced with a lot of problems when started working on a nics table.

 

[dcsapak]

hi,
yes i use grafana for a dashboard,
here the proxmox relevant graphs i use:

tried to create several dashboards, but faced with a lot of problems when started working on a nics table.

what exactly did you try, and what did not work?

 

[nicolaiB]

We also use influx with proxmox. Unfortunately the data model (esp. units for measurements) isn't documented, so I had to look into the source code for reverse engineering. I'm planning to update the wiki page.

 

[dcsapak]

I'm planning to update the wiki page.

great, looking forward to this thanks

 

[Melanxolik]

We also use influx with proxmox. Unfortunately the data model (esp. units for measurements) isn't documented, so I had to look into the source code for reverse engineering. I'm planning to update the wiki page.

Look really cool, could you please share your dashboard? I tried use this query for build graph for net:

SELECT mean("netout") FROM "nics" WHERE object = 'qemu' and  netout > 9000000000 and $timeFilter GROUP BY time($__interval),host,vmid fill(null)

but it had problem with timedelta, after that I changed query to:

SELECT DERIVATIVE("netin") FROM "nics" WHERE "vmid" = '$vmid'  and $timeFilter

but it didn't provide me good info too....

 

[nicolaiB]

Attached you find the dashboard. You have to remove the .txt suffix.

Regards, Nicolai

 

[Melanxolik]

I am really appriciated, you did good job.

 

[Melanxolik]

Look cool, but one question, does it possible to show instance name near vmid on the right?

When I used query:

SELECT mean("netout") FROM "nics" WHERE object = 'qemu' and  netout > 9000000000 and $timeFilter GROUP BY time($__interval),host,vmid fill(null)

it gave me detail information about virtual instance name

 

[Melanxolik]

hi,
yes i use grafana for a dashboard,
here the proxmox relevant graphs i use:

what exactly did you try, and what did not work?

Cool dashboard too, I just tried create something similar but I got problem with graph...

My graph for avg* has a lot of points, but doesn't have line. If I turn off option: Points, my graph will be empty without any lines. I can't undestand why.

 

[nicolaiB]

Look cool, but one question, does it possible to show instance name near vmid on the right?

When I used query:

SELECT mean("netout") FROM "nics" WHERE object = 'qemu' and  netout > 9000000000 and $timeFilter GROUP BY time($__interval),host,vmid fill(null)

it gave me detail information about virtual instance name

Please find attached to this post a enhanced version.

N

 

[nicolaiB]

Cool dashboard too, I just tried create something similar but I got problem with graph...

My graph for avg* has a lot of points, but doesn't have line. If I turn off option: Points, my graph will be empty without any lines. I can't undestand why.

Sound like a problem with derivation / interval. Could you share your dashboard, so could I take a look.

 

[Melanxolik]

yes, please. It will be cool if you can find problem

 

[nicolaiB]

Simple solution: change 'Null value' to 'connected' (and change Query A measurement to load 1 - looks like a typo)

 

[czechsys]

Oh, thank you very very much for this, i was struck some months ago with this...

 

[Melanxolik]

nicalaiB
sorry, I didn't get your advice clear: (and change Query A measurement to load 1 - looks like a typo)

SELECT mean("avg5") FROM "cpustat" WHERE "host" =~ /^$node$/ and $timeFilter GROUP BY time($__interval),"avg5","host" fill(null)

could you please provide some example ?

 

[nicolaiB]

Sorry should be avg1 and not load1

 

[Melanxolik]

Sorry, probably I didn't explain my question so clear, I just wanted ask about: How can I change description of parameter on the graph.
Example: from cpustat.mean {avg1:, cluster-3-1:} to simple: avg1.
I couldn't find this way to do it.

And if you see on the graph which show: dcsapak, he didn't have any label with world "mean"

Thank you for your helping.

 

[dcsapak]

this is done by the alias function for single values, or the alias field for a whole query

 

[Melanxolik]

Sorry guys, my problem with description was solved on last version graphana.

 

[Melanxolik]

If you are using lastet version grafana you can do query:

SELECT derivative(mean("wr_bytes"), 1s) AS "$vmid" FROM "blockstat" WHERE "nodename" =~ /^$node$/ AND "vmid" =~ /^$vmid$/ AND $timeFilter GROUP BY time($__interval), "vmid","host" fill(null)

and filed: Alias BY: $tag_vmid $tag_host

After that you will have VMID +VMNAME on charts.