Warning and critical lines in Highcharts Info

[WillemDH]

Hey,

Just noted that the warning and critical lines are not supported with some of my plugins. I'm guessing I need to adjust my perfdata output and add the warning and critical thresholds.

Before I start working on this, I would need to be sure it's going to work for some of them. Let's take for example the check_ms_win_disk_load plugin. As it has 8 different perfdata sources, where there can be warning and critical values for read queue and write queue, I'm not sure if this would even work...? I'm guessing you can't have 2 warning and critical lines..?

Thoughts?

Grtz

Willem

[jdalrymple]

Willem,

we noticed this in testing. I can't remember if swilkerson asked for me to create a bug report for it or if he was just going to take the ball and run. I'll catch up with him about it later today though.

The warning and critical lines aren't showing for thresholds that have a range. So basically if it's anything besides a horizontal rule (noting that the default behavior is for values above a certain threshold) the warning/critical markers aren't showing up.

I'd want to verify this in my lab, which is still napping this fine Monday morning, but I'm almost certain that was the behavior we experienced, and as mentioned I'll make sure that either swilkerson is on top of the problem or that we have a bug report so it doesn't get lost in the fray.

[WillemDH]

Hey Jdalrymple (still don't know your first name? => please tell me, can't find it on https://www.nagios.org/about/team/),

Thanks for looking into this. It would be nice to know what the expected behavior should be. I also noticed that for example with the check_xi_host_ping, only the warning lines for the pl are shown, not those for the thresholds of the rta.

I can imagine this is a difficult subject and it is plausible to only show one warning and one critical line. But how is it determined which perfdata is taken when there are multiple options...

Grtz

Willem

[jdalrymple]

Hi Willem,

I'm JR, and wow is that webpage out of date. I'll file a bug report for the marketing department to deal with that

I forgot that swilkerson is out. It will be a few days before I have definite answers as a result. I'm thinking he probably didn't do anything yet as that was a very last minute find during conference prep.

[WillemDH]

JR,

Any new on this topic yet? We also noticed that when we change the warning / critical thresholds in the CCM, the Highcharts graphs are not getting updated, except when we delete the rrd file after which it is recreated with the correct threshold. is there any way to automate this?

Grtz

Willem

[jdalrymple]

There is a bug report in. I talked to scottwilkerson right after he returned but didn't get the bug in until today for tracking purposes. Honestly what we came up with is that there isn't an obviously great solution to a number of the problems. The problems are:

1) inverse thresholds (things like uptime) don't work - this is easy to fix

2) range thresholds don't show up - this isn't easy to fix because rrdtools isn't equipped to handle it

3) only 1 set of thresholds on a graph with many metrics - I suggested that we omit the thresholds on the main performance graph and offer a button to blow up all the metrics into their own graphs each with a threshold. That will likely be complicated though so I wouldn't expect that in a big hurry, I would guess they will just omit the thresholds for now.

4) your new problem - I don't know how hard or easy this is. I suspect there is a command they can run to relocate the threshold lines in the RRD without blowing it up though, so fingers crossed.

Did I miss anything?

[WillemDH]

Didn't know about 1 and 2.

For 3, do you happen to know which set of thresholds will be shown? A button to blow up all metrics into their own graph each with their own threshold would be an awesome addition to Nagios! +1 from me.

4 => Can we keep this thread open until someone comes up with such a command to update the rrd with the new threshold?

Tx!

[jdalrymple]

For 3, do you happen to know which set of thresholds will be shown?

Any answer I gave I wouldn't trust. I can't correlate it - honestly. That's why it needs to just get chucked.

A button to blow up all metrics into their own graph each with their own threshold would be an awesome addition to Nagios! +1 from me.

That was my idea I'll take it as a compliment

Can we keep this thread open until someone comes up with such a command to update the rrd with the new threshold?

Only if you don't answer - our dashboard thing...

[WillemDH]

Had the same problem again. The warning and critical threshold line suddenly changed after editing the warning threshold in nagios.. This is causing weird results.

For example:
First I had this as output:

Code: Select all

'php_cpu'=0%;80;90;0;100 'php_mem'=1%;80;90;0;100 'php_count'=5;1;25;0;99999

Warning and critical threshold were 80 and 90 in the Highcharts graphs.

Then I changed the warning and critical threshold to 40 and 60 (in Nagiios)

Code: Select all

'php_cpu'=0%;40;60;0;100 'php_mem'=1%;40;60;0;100 'php_count'=5;1;25;0;99999

And suddenly Highcharts did no longer show 80 and 90, nor 40 and 60, but 1 and 25, which are the warn and critical threshold of process count. This is really confuing and not consistant. For me it would be ok if Highcharts would always take the first warning / critical threshold in the list. But now it's like he takes the first, but when the warn / crit values change, it seems to take another perfdata's value (the last one or the first one that is not changed?)

The weird thing is that after I tried deleting the rrd and xml of this service, it again took the last value (count).

I think the solution JR proposed to omit all warning / critical threshold lines for services containing multiple perfdata metrics and a offer a button to blow up all the metrics into their own graphs each with a threshold would be the best solution.

The warning and critcal threshold lines are only confusing atm for services containing multiple perfdata. Maybe it would be a good idea to give us an option to enable the warning / critical threshold lines, but omit them for services containing multiple perfdata? This awaiting the better solution JR propsed (which I understand could take a little longer to implement? It's just that right now it all seems very inconsistent. (If I have 5 services each monitoring one process and the warning / critical threshold is in every graph based on a different metric is different....)

Grtz

Willem

[tmcdonald]

I tend to agree, and I think that would make a good feature request. I'll make sure that gets filed and will update this thread once this is done (or have whoever files it do so).