I'd like to use a sparkline for quick volume context in conjunction with a tstats command because of its speed. If the following works
index=foo | stats sparkline
I would have assumed this would work as well
| tstats count where index=foo by _time | stats sparkline
I've tried a few variations of the tstats command. In the end what I generally get is a straight line which I'm interpreting to mean it is showing me there is a 'count' event for that time. If I use span in the tstats 'by' command the straight line becomes jagged but consistently so.
If you specify the span in tstats, as well as adding specifics to the sparkline command, I think you'll have better luck. Here is an example search for the past 24 hours:
| tstats count where index=foo by _time span=10m | stats sparkline(sum(count), 10m) as Volume
Because | tstats returns tabled results instead of events, the default for sparkline just shows a count of 1 for each row. By specifying a sum for count, you should get better results.
If you already have your tstats command outputting the format of events you want, you can just tack on this to the tstats command if you don't already have a BY clause:
by _time span=10m
| eventstats sparkline(sum(count), 10m)
| fields - _time
If you already do have a BY clause, just tack on the _time field and span argument. Then your results will have a sparkline just like as above, except that you don't need to change your tstats to use prestats or any other changes to avoid a double-transforming command.
This is what I needed for making a sparkline on a Single Value visualization. Worked very well!
If you specify the span in tstats, as well as adding specifics to the sparkline command, I think you'll have better luck. Here is an example search for the past 24 hours:
| tstats count where index=foo by _time span=10m | stats sparkline(sum(count), 10m) as Volume
Because | tstats returns tabled results instead of events, the default for sparkline just shows a count of 1 for each row. By specifying a sum for count, you should get better results.
The sparkline(sum(count)) bit was the trick! Adding the span and time piece in the sparkline arguments isn't specifically needed though certainly they have their place. Thanks.
No problem, glad I could help! I like adding the time elements in to make sure I get what I'm expecting, but certainly not required.
Do you see anything different if you add span=1h
or some other relevant span increment to the tstats
search command?