InfluxDB features a SQL like query language, only used for querying data. The HTTP API has endpoints for writing data and performing other database administration tasks. The only exception to this is continuous queries, which perpetually write their results into one or more time series.
Getting a List of Time Series
There are two different methods for returning a list of all time series in a database:
list series -- or this select * from /.*/ limit 1
The first query will return all series, while the second will return the most recent point from each series that matches the given regex.
Select and Time Ranges
By default, InfluxDB returns data in time descending order. The most efficient queries run over only a single column in a given time series.
select value from response_times;
This simple query pulls the values for the
value column from the
How to set query start and end time
If start and end times aren’t set they will default to beginning of time until now, respectively.
time is built in for every time series in the
You specify the start and end times by setting conditions on
time columns in the where clause.
Below are the different formats that can be used to specify start and end times.
Date time strings
Date time strings have the format
YYYY-MM-DD HH:MM:SS.mmm where
mmm are the milliseconds within the second.
select value from response_times where time > '2013-08-12 23:32:01.232' and time < '2013-08-13';
The time and date should be wrapped in single quotes.
If you only
specify the date, the time will be set to
the hours, minutes, and seconds is optional and specifies the
You can use
now() to calculate a timestamp relative to the server’s
select value from response_times where time > now() - 1h limit 1000;
will return all points starting an hour ago until now.
Other options for how to specify time durations are
s for seconds,
m for minutes,
h for hours,
for days and
w for weeks.
If no suffix is given the value is
interpreted as microseconds.
You can specify timestamp in epoch time, which is defined as the number of microseconds that have elapsed since 00:00:00 Coordinated Universal Time (UTC), Thursday, 1 January 1970. You can use the same suffixes from the previous section if you don’t want to specify timestamp in microseconds. For example:
select value from response_times where time > 1388534400s
will return all points that were written after
Selecting a Specific Point
Points are uniquely identified by the time series they appear in, the time, and the sequence number. Here’s a query that returns a specific point.
select * from events where time = 1400497861762723 and sequence_number = 2321;
Note that the time is a very large number. That’s because it’s a microsecond scale epoch. InfluxDB always stores points at this scale, but most libraries will return the time as either a second, or millisecond scale value. If you’re selecting a specific point, you’ll need to know the exact microsecond scale epoch that point has otherwise you’ll get an unexpected empty result.
Selecting Multiple Series
You can select from multiple series by specifying a regex to match against. Here are a few examples.
select * from /^stats\./i where time > now() - 1h;
Get the last hour of data from every time series that starts with
stats. (case insensitive).
select * from /.*/ limit 1;
Return the last point from every time series in the database.
Deleting data or dropping series
The delete query looks like the following:
delete from response_times where time < now() - 1h
With no time constraints this query will delete every point in the
You must be a cluster or database admin to run delete queries.
You can also delete from any series that matches a regex:
delete from /^stats.*/ where time < now() - 7d
Any conditions in the where clause that don’t set the start and/or end time will be ignored, for example the following query returns an error:
delete from response_times where user = 'foo'
Deleting all data for a series will only remove the points.
It will still remain in the index.
If you want to remove all data from a series and remove it from the list of series in a database use the
drop series response_times
Note about Delete Performance
Currently, deletes are not very efficient. If you want to quickly evict old data, the best way to do that is by dropping a shard. For more information on shards go here.
The Where Clause
We’ve already seen the where clause for selecting time ranges and a specific point. You can also use it to filter based on given values, comparators, or regexes. Here are some examples of different ways to use where.
select * from events where state = 'NY'; select * from log_lines where line =~ /error/i; select * from events where customer_id = 23 and type = 'click'; select * from response_times where value > 500; select * from events where email !~ /.*gmail.*/; select * from nagios_checks where status <> 0; select * from events where signed_in = false; select * from events where (email =~ /.*gmail.*/ or email =~ /.*yahoo.*/) and state = 'ny';
The where clause supports comparisons against regexes, strings, booleans, floats, integers, and the times listed before.
= equal to,
> greater than,
< less than,
<> not equal to,
=~ matches against,
!~ doesn’t match against.
You can chain logic together using
or and you can separate using
The group by clause in InfluxDB is used not only for grouping by given values, but also for grouping by given time buckets.
You’ll always be pairing this up with a function in the
Here are a few examples to illustrate how group by works.
-- count of events in 10 minute intervals select count(type) from events group by time(10m); -- count of each unique type of event in 10 minute intervals select count(type) from events group by time(10m), type; -- 95th percentile of response times in 30 second intervals select percentile(value, 95) from response_times group by time(30s);
By default functions will output a column that have the same name as the function, e.g.
count will output a column with the name
count in order to change the name of the column an
AS clause is required.
Here is an example to illustrate how aliasing work:
select count(type) as number_of_types group by time(10m);
The time function takes the time interval which can be in
microseconds, seconds, minutes, hours, days or weeks.
To specify the
units you can use the respective suffix
Filling intervals with no data
By default, group by intervals that have no data will not have associated datapoints. For instance, say you have the following query:
select count(type) from events group by time(1h) where time > now() - 3h
If the events series had data for this hour and two hours ago only, you’d only get two points in the result.
If you want to ensure that you get back points for intervals that don’t have data, you can use the
fill function like this:
select count(type) from events group by time(1h) fill(0) where time > now() - 3h
fill must go at the end of the group by clause if there are other arguments:
select count(type) from events group by time(1h), type fill(0) where time > now() - 3h
You can merge multiple time series into a single stream in the select clause. This is helpful when you want to run a function over one of the columns with an associated group by time clause.
select count(type) from user_events merge admin_events group by time(10m)
You’d get a single time series with the count of events from the two combined in 10 minute intervals.
Joins will put two or more series together. Since timestamps may not match exactly, InfluxDB will make a best effort to put points together. Joins are used when you want to perform a transformation of one time series against another. Here are a few examples.
select hosta.value + hostb.value from cpu_load as hosta inner join cpu_load as hostb where hosta.host = 'hosta.influxdb.orb' and hostb.host = 'hostb.influxdb.org';
The above query will return a time series of the combined cpu load for hosts a and b. The individual points will be coerced into the closest time frames to match up.
select errors_per_minute.value / page_views_per_minute.value from errors_per_minute inner join page_views_per_minute
The above query will return the error rate per minute.