Point tags are key-value pairs (strings) that are associated with a point. Point tags provide additional context for your data and allow you to fine-tune your queries so the output shows just what you need.
Note: If point tag values are blank or empty, errors can result. Point tag values can be zero.
Point Tag Basics
Point tags offer a powerful way of labeling data so that you can slice and dice it in almost any way you can imagine. For example, you can use point tags, to label a point’s datacenter, version, etc. and can then group by datacenter or version.
Many of our cloud integrations generate point tags automatically to help you filter metrics. You add point tags explicitly using Wavefront proxy preprocessor rules.
Point Tag Maximum
Wavefront has limited the number of point tags to 20 for most clusters. Our experience has shown that a larger number of point tags does not improve the user experience, and can lead to performance problems.
Note: If the number of point tags exceeds 20, then we drop the metrics that have those point tags.
Point Tag Example
Our ~sample metrics, included in each Wavefront instance, have two point tags,
az. Each point tag has two values, and you can filter to show one or the other set of time series. Below, the time series with
env=dev are shown in shades of blue, and the time series with
env=production are shown in shades of green.
You can filter further, either by specifying another point tag or by combining a filter for a point tag filter with a source filter.
Finally, instead of filtering, you can use grouping, for example, in conjunction with an aggregation function. The sample data are a bit unusual because all time series with
env=production are also tagged with
az=us-west2. In an actual data set, we would expect 4 lines as a result of this query, two that sum for each environment and two that sum for each availability zone.
Best Practices for Point Tags
To avoid performance issues, follow best practices.
Enclose Point Tag Values in Double Quotes
Double quotes are required if the point tag value includes certian characters such as spaces, but are recommended in all cases. For example, use
"my test" instead of
Don’t Use More Than 20 Point Tag Keys
Each point tag key (e.g.
az) can be associated with a high number of values, but you can’t have more than 20. You’ll find that working with your data shape to fall inside this limit has the side benefit of making it much easier to understand what you see.
Watch the Number of Time Series
Wavefront recommends that you keep the number of distinct time series per metric and host to under 1000. Whether a time series is distinct depends on the combination of the point tag keys and the point tag values.
For example, assume a metric
cpu.idle and a host
web1. If you use that metric and host with the point tags
datacenter=atl, a new time series results. If you use
datacenter=atl, another distinct time series results.
Don’t Use Point Tags for Highly Variable Data
Using point tags to store highly variable data such as timestamps, login emails, or web session IDs will eventually cause performance issues when your data are queried. That is also true if you specify a time that results in many time series being retrieved. For example
timestamp=<now> or even
monthofyear=11 can exceed the limit. In contrast,
monthofyear=jan are acceptable.