topic Re: Scale Bucket size for time series data in Deployment Architecture

Scale Bucket size for time series data

bondu — Sun, 19 Oct 2014 17:47:28 GMT

I am trying to generate a chart where the x-axis scales over time. So the columns would be:
1 hour, 1 day, 1 week, 1 month, 3 months, 6 months, 1 year

Do I need to create each bucket individually, or can I do it with some sort of log scale? The columns don't need to be exactly the date ranges above.

Re: Scale Bucket size for time series data

martin_mueller — Sun, 19 Oct 2014 22:56:20 GMT

Assuming you want to look back from the end of the time range into the past, here's an idea:

  index=_internal | addinfo | eval diff = info_max_time - _time
| eval bucket = case(diff < 60, "1m", diff < 3600, "1h", diff < 86400, "1d") | chart count by bucket

I'm grabbing the end of the time range to calculate how "old" an event is, then I'm shoving the events into custom buckets and charting by that.

Re: Scale Bucket size for time series data

gkanapathy — Sun, 19 Oct 2014 23:35:47 GMT

You can also (if you want) try something like:

... | addinfo | eval diff= info_max_time - _time | bucket span=1log10 diff | chart count by diff

You can experiment with different values for the span, e.g., 1log2 or 1.2log10. This will give you time differences in seconds. The number of seconds might be awkward to work with though, so the method above with case() would be better in that case.

Re: Scale Bucket size for time series data

bondu — Mon, 20 Oct 2014 19:12:46 GMT

I was able to use this example with a few changes:

index=_internal | addinfo | eval diff = info_max_time - _time | eval bucket = case(diff <= 86400, "1 day", 86400 < diff AND diff <= 172800, "2 days", 172800 < diff AND diff <= 604800, "1 week", 604800 < diff AND diff <= 1209600, "2 weeks", 1209600 < diff AND diff <= 2628000, "1 month") | chart count by bucket

Re: Scale Bucket size for time series data

martin_mueller — Mon, 20 Oct 2014 22:33:49 GMT

Great... I'll mark this as solved?