Nagios Support Forum

I have been monitoring an Exchange host and I currently have 64 different checks (mostly perfmon counters).

Each service has been configured to create a performance graph.

I go to the Host Status Detail page and click on the Performance Graphs tab.

This then starts to display all the 64 graphs ... a little slowly but they eventually all appear. The screen will then be shortly updated.

When I was working on a client site remotely I was connected via a Microsoft PPTP VPN tunnel. I was using my web browser to connect to Nagios XI across the VPN link.

When I went to view all 64 graphs on the Performance Graphs tab it never actually finishes downloading all the graphs. It does about 40 and then starts all over again.

At this point I happened to look at the performance of the Nagios XI VM CPU usage on my ESX host.
Area 1
This is the normal CPU usage of the Nagios XI host. In total it is performing 93 checks, 64 of them every 1 minute and the remainder ever 5 minutes. I am OK with this CPU usage.

Area 2
This is the CPU usage when I went to view all 64 graphs over a VPN.

Area 3
Leaving my existing browser window open, trying to load these graphs, I RDP'd to a server onsite and opened a browser inside the RDP session. All the graphs loaded as expected, a little slowly. Leaving this browser open and the broswer going across the VPN caused this cpu usage (I might have closed a browser and opened again, this would be the drop).

Area 4
This is when on both browsers, I nagivated elsewhere and then retured to the Host Status Detail tab, but DID NOT click the Performance Graphs tab. I have noticed that when the page initially loads I can see it loading all the dashlets for the graphs but then they are hidden. So it seems there is still quite a bit of activity going on in the background when the graphs are not even being viewed.


Suggestion
Perhaps limit the amount of graphs that can be displayed at once, break them up into pages of 10 or something like that. Perhaps a user customisable setting depending on the performance of their server.
My only other idea is if it was possible to create multiple service groups and break the 64 services up, this way I can see the graphs for each of these service groups without overloading the server.


I think for the time being I'll copy the existing host definition and give it a different name but pointing to the same host. Then I'll go and split the 64 services across three or four of these copies and change the "config name" on the services to match up. This should let me see the graphs on one page without having to create a custom dashboard view.

*Edit* Actually it's only 50 active services, 14 of them are disabled.

I'm compiling a list of these issues. I'm not sure if there will be a solution, as I don't feel we will be maintaining a Web1.0 interface for slow bandwidth links. We will try and balance usefulness and robustness, but I fear that features, bells, and whistles will win out over a light weight interface.

The increased CPU is due to RRDTool rendering all those graphs for you. The reason you only got 40 images on a slow link before they started over, is that the graph images automatically refresh every 30 or 60 seconds (I don't recall exactly). The refresh time is counted from the time the image is called (the fetch is started), rather than when the image download is complete. I don't know if there's a javascript method for detection completion time of an image download.

Accessing Nagios XI over a WAN link to view performance graphs isn't a major thing. I think it was interesting to see how this problem had a somewhat chain-reaction affect on the Nagios XI server CPU usage.

I do think that there needs to be some sort of solution that allows you to view performance graphs of a host on a page by page basis where each page displays say 10 graphs per page (user definable). Specifically as you go from page to page it is at this point that RRDTool should be generating the graphs, not doing all the graphs in the background for all the pages as this is where the CPU usage goes up.

One possible solution is to put these images into the HTML using data URLs, this way the page won't load until the image data is fully downloaded.

http://en.wikipedia.org/wiki/Data_URI_scheme

As this uses base64 encoded images implementing transparent compression would be beneficial to save time/bandwidth.

R1.2 Release Notes wrote:Added paging options to host service performance graphs pages (default to 5 records at a time)

Thanks for the changes implemented in the new release, they're great.

With R1.2

I am looking at a host that has more than 5 graphs.
I click the More Performance Graphs link
I change the number of entries per page from 5 to another number and click Go.
The screen refreshes and now shows all performance graphs on the Nagios XI server.

That is all graphs for all hosts services.

Thanks for getting this fixed, works great now.