Nagios Conference 2012 - Anders Haal - Why dynamic and adaptive thresholds matters
0 likes1,226 views
Anders Haal's presentation on using adaptive thresholds with Nagios. The presentation was given during the Nagios World Conference North America held Sept 25-28th, 2012 in Saint Paul, MN. For more information on the conference (including photos and videos), visit: http://go.nagios.com/nwcna
![....
<! 12:00 Static >
<hour>7000</hour>
<! 13:00 Adaptive >
<hour>erpserverordersediOrders[0] / 3</hour>
....](https://image.slidesharecdn.com/andershaalwhydynamicandadaptivethresholdsmatters-121003150701-phpapp01/85/Nagios-Conference-2012-Anders-Haal-Why-dynamic-and-adaptive-thresholds-matters-30-320.jpg)
![....
<! 12:00 Static >
<hour>7000</hour>
<! 13:00 Adaptive >
<hour>erpserverordersediOrders[0] / 3</hour>
<! 14:00 Adaptive with math function >
<hour>avg(erpserverordersediOrders[30M:60M]) / 2</hour>
....](https://image.slidesharecdn.com/andershaalwhydynamicandadaptivethresholdsmatters-121003150701-phpapp01/85/Nagios-Conference-2012-Anders-Haal-Why-dynamic-and-adaptive-thresholds-matters-31-320.jpg)
![Infrastructure monitoring
Application performance monitoring [APM]
Business activity monitoring [BAM]
Operational Business intelligence [OBI]](https://image.slidesharecdn.com/andershaalwhydynamicandadaptivethresholdsmatters-121003150701-phpapp01/85/Nagios-Conference-2012-Anders-Haal-Why-dynamic-and-adaptive-thresholds-matters-36-320.jpg)
Ad
More Related Content
What's hot (20)
Similar to Nagios Conference 2012 - Anders Haal - Why dynamic and adaptive thresholds matters (20)
Ad
More from Nagios (20)
Ad
Recently uploaded (20)
Nagios Conference 2012 - Anders Haal - Why dynamic and adaptive thresholds matters
- 3. Bischeck dynamic & adaptive thresholds for Nagios www.bischeck.org
- 4. Threshold
- 5. What is the limitation with static threshold? ✗ Not static ✗ Load varies throughout the day, week ✗ To many or to few alarms ✗ Collecting and thresholding in the same context ✗ Based on the current measurement ✗ Do not consider dependency to other services
- 8. {example 1} “Database table size should not be bigger then 5 % of max size yesterday“ 200000 180000 160000 140000 table size < max(yesterday)*1.05 120000 100000 Table size 80000 60000 40000 20000 0 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Yesterday Today
- 10. {example 2} “Number of online users should not be more then 10 % higher then the average number of online users for the last 10 data points” users < avg(X0+X1+.....+X9)*1.1 Where X is the historical on-line users data points
- 12. {example 3} “The number of orders with errors should be lower then 5% of the total number of registered orders” Total #orders #orders with errors < 0.05 * (Total #orders)
- 13. {example 4} “Message queue size should be above the defined Friday threshold profile” count time of the day
- 14. {example 4} “Message queue size should be above the defined Friday threshold profile” count count time of the day time of the day
- 15. How to make thresholds dynamic & adaptive? ✗ Time profiles ✗ Historical data points ✗ Math and statistical operations
- 16. We did not want a check_XYZ hack We wanted a tool
- 17. Collecting
- 18. Collecting Separation Thresholding
- 20. Collecting Logic Historical data
- 21. Collecting Day profile Logic Historical data
- 22. Collecting Day profile Calender Logic Historical data
- 23. Collecting Day profile Calender Logic Historical data Nagios
- 24. Collecting Scheduling Day profile Calender Logic Historical data Server Interface OpenTSDB Nagios XYZ
- 27. bischeck basics ● Configuration like Nagios – host, service but also service item ● Host is just a container of the rest ● Service specify the connection and scheduling ● Service item specify the “query” and the threshold class to use ● Host and service name must be the same as in the Nagios configuration
- 28. Threshold – 24 hour day profile ● Divide the day in 24 hour points, where every point can be: ● Static value ● Dynamic value – Math expression on single value or range of data from the cache – Based on cached data points retrieved by ● Index – single value or index range ● Time – single value (closest) or time range (between)
- 32. Threshold – 24 hour day profile Between every “full” hour a linear equation is calculated Day profile 60000 50000 40000 30000 20000 10000 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Hour
- 33. Threshold – 24 hour day profile ● Connect calender to the day profile and evaluate according to the following order: 1. Month and day of month 2.Week and day of week 3.Day in month 4.Day in the week 5.Month 6.Week ● Holiday – exception days
- 34. And more.... ● Multithreaded and multischeduling schema per service ● interval ● cron ● Data collection – jdbc, livestatus, internal cache ● Virtual services ● Date macros in execution statements ● Customize ● connection (service classes) ● execution (service item classes) ● thresholds (threshold classes) ● server integration (server classes) ● XML configuration supported with WEBui (beta) ● GPL 2 license
- 35. Future ● Improved time series database ● Patterns/baselines ● More statistic functions ● “Sensors” alarms on multiple/aggregated data points ● Any ideas?