LMS 4.0 IP SLA Availablity Report is Incorrect

[xcz504d1114]

I'm polling a few thousand locations using IP SLA, I have responder enabled on all destinations, and I'm using 60 byte voice packets with a QoS policy.

When I run an IP SLA Summary availability report, I have a bunch of locations showing 9% availability 8.5% etc. When I go to the actual collector, and pull up a graph of the same time period, that graph shows 100% availability.

Same collector, same data, just different views giving completely different results. I have to assume that the IP SLA summary report is wrong, these sites were not down 90% of the time.

Just a random though to go with that, I do have the IP SLA to only pull information during the locations operational hours, and I did pull the report from midnight to 11am, the statistics should have been gathered for 4 hours of the 11, which is still higher than 9%, and I would expect all of my locations to report like that, not just a few hundred.

All of the devices are similar in hardware and IOS, and I have verified on a handful that IP SLA responder is enabled, and I see the connections, I have also verified the source configuration via command line.

Any thoughts?

Thanks,

Craig

[Joe Clarke]

What's the report period instance type?  If you're doing an extended period of time, perhaps these collectors really weren't available at some point over the time period.  If you select a report period if latest polled, is this 100%?

[xcz504d1114]

If I pull the latest polled, it says that there is nothing to report on. The collector manages 1000+ SLA connections, when i look at the specific ones that show 0% availability on the collector, it shows good success and the last operation was ok, the responder shows the communication as well.

When looking at the graph from the collector page, for the exact same time period, it shows what I believe is a proper up time, and the report generator shows 0%.

I think I might have to call TAC, this seems more like a bug.

[xcz504d1114]

The detailed report also shows correct information, this seems to only be effecting the IP SLA summary availability report.

[xcz504d1114]

There is something with the way it is reading teh database when applying "Report Ranges". If I run a report on just the one location, with a range of 1% to 100%, it shows as 99.86% available, when I run the exact same report with 0% to 100% it shows up as 66.58% available, and when I run it from 0% to 99.5%, it shows as 0% available.

I'm not sure why changing the report range effects availability, but 0 - 100 seems to be the most accurate, at least for that one location, unfortunately that makes the output of my report huge, and the reliability of the data is obviously questionable.

[Joe Clarke]

You're absolutely right.  The way the query works is that it applies the from and to ranges to the raw data and not the resulting averages.  The result is that the average is only applied to those entries that fall between the from and to ranges.  This means that the more narrow you make the range, the smaller the resulting availability will be.  I have a patch that corrects this.  If you open a TAC service request I can provide you the patch.  Just use the Open Service Request link in the action panel of this thread.