This is an internal project my team developed @Progelspa, the idea is very simple: leverage Log Analytics data to create a shortlist of usage trend patterns that may need to be investigated by a human. The solution addresses the following requirements:
- be able to specify a set of intervals to compare the usage with. A common setting we use is: compare data from last week, to the previous week and to the previous month.
- be able to specify an absolute attention threshold
- be able to manage both growing and decreasing counters, for example processor usage and disk free space
- be able to specify an ignore threshold to avoid trending on under utilized systems
- be able to use any aggregation function available in OMS
- be able to send the shortlist by email using visual sparklines to help the human understand at a glance
- and obviously be able to specify the notable differences in percentage or absolute value
Indeed this is what we have achieved so far, it is working and we’re using it, but as far as appetite grows with eating there are a ton of things that can be added, for example with Log Analytics we can have aggregation on any data type not only for performance data, a lot needs to be done in terms of error checking and so on.
So the question for the community is: is this something valuable? Anyone available for contributions?
I don’t even know if will be cleared for open sourcing the project, but before any attempt I need to check if it is worth or not.