Skip to content

Latest commit

 

History

History
90 lines (61 loc) · 2.3 KB

README.md

File metadata and controls

90 lines (61 loc) · 2.3 KB
Raw

InfluxDB Fetcher

About

Use this program to fetch data from InfluxDB using the HTTP query API and export it into InfluxDB line protocol format.

Synopsis

influxdb-fetcher uri username password database query [fieldstotags]

Example

influxdb-fetcher \
    http://influxdb.example.org:8086 root root aqi_readings \
    "SELECT station_type, latitude, longitude, aqi_value FROM testdrive GROUP BY station_type LIMIT 100"

Setup

wget --no-clobber --output-document=/usr/local/bin/influxdb-fetcher https://raw.githubusercontent.com/hgomez/influxdb/master/bin/influxdb-fetcher
chmod +x /usr/local/bin/influxdb-fetcher

Advanced usage

Transform fields to tags

You could also transform fields into tags by adding them as a list of comma- separated labels as a last parameter to the command line.

The background of this is that since as of today, there is now way to transform fields into tags. See also: influxdata/influxdb#3904.

In the following example, latitude and longitude fields will be put as tags into the line protocol format output.

influxdb-fetcher \
    http://influxdb.example.org:8086 root root aqi_readings \
    "SELECT * FROM testdrive GROUP BY station_type LIMIT 100" \
    latitude,longitude

Rename fields

As the line protocol format is pure ASCII, it is easy to use standard Unix tools like sed to manipulate the content.

# Fetch data.
influxdb-fetcher ... > data.wireproto

# Manipulate schema: Rename field.
sed -i -e "s/foo\=\([0-9.]*\)/bar=\1/g" data.wireproto

# Manipulate data: Cast from Int64 to Float.
sed -i -e "s/value\=\([0-9]*\)i/value=\1/g" data.wireproto

# Upload data to different destination.
curl -u login:password -i -POST \
    "http://dest-influxdb.example.org:8086/write?db=new_db" \
    --data-binary @data.wireproto

Notes

  • To get InfluxDB tags properly populated into line protocol format, you should add a GROUP BY clause.

  • By carefully crafting InfluxQL expressions (using WHERE time=...), you should be able to fetch data in batches, ideally 10K-20K at a time. This way you can extract chunks of the source data. The background of this is that InfluxDB will OOM with too large POST requests.

Development

Build code

cd influxdb-fetcher
mvn clean package