Odata provider (server) interfaces influxdb by exposing using REST based odata endpoints. Supports Influxdb 1.x and 2.x Connects InfluxDB with BI apps including Excel, PowerBI and Tableau.
Current version has been tested on python 3.6. Few features might break in 2.x.
Install all dependencies via pip
pip install -r requirements.txt
Configure Odata server host and port details in
settings.conf Sample config file is well documented. Choose the
influxdb_version in metadata section and accordingly provide
influxdb connection details. Also, enable authentication if required.
Refer to Authentication section
Before spinning up the server, create an xml metadata defining the fields/tags of each measurement in an influx database (or bucket) using
python startServer.py -c production.conf -m
-m or --createmetadata flag creates Odata schema metadata file
at the location defined by the metadata_file tag in the
configuration. It doesn't spin up the server. Schema is generated
for all measurements specified by the databases or bucket tags
in the configuration. Multiple db/buckets can be configured by
separating each by comma(,)
Once the metadata file is generated, spin up the Odata server by
python startServer.py -c settings.conf
You may wish to install the influx-odata.service file and run this on system start. Note that the file currently is configured for use with NGINX, python3.8, and Centos8, so modify accordingly.
If you use SELinux, you will need to enable the httpd_can_network_connect
and nis_enabled booleans. You may also need to adjust the labels and
generate some policy modules. In my case, I needed policies for
transition, search, nnp_transition, and entrypoint before Python
was able to start under Systemd. A similar hurdle likely exists for
AppArmor.
- Odata 2 provider (server) implemetation.
- Supports both Influxdb 1.x and Influx 2.0 with Flux support. Based on the request, it internally converts into influxql/flux query (based on influx version in use).
- Works with any BI tool supporting Odata 2 as data source. Tested for Excel 2016, 2019, Office 365, Tableau Desktop 2019.
- Supports only one database/bucket per service. Multiple instances on separate ports can be initiated to handle multiple databas/bucket.
- Database/Bucket and Measuremnt
- Select
- Filter (time based and tag-value based)
- Aggregate Function
- Group by time
- Limit
For example, the endpoint
http://localhost:8080/data/monitoring__cpu?$filter=timestamp ge datetime'2020-04-03T00:00:00' and timestamp lt datetime'2020-04-06T00:00:00' and host eq '192.168.1.12'&$top=100&groupByTime=1h&aggregate=mean&$select=usage_guest,usage_guest_nice,usage_idle,usage_iowait, usage_steal,usage_system,usage_user
queries measurement cpu inmonitoring database for timestamp
greater equal to 2020-04-03T00:00:00 and less than 2020-04-06T00:00:00
and tag host equal to string value 192.168.1.12, group by
every 1 hour and returns aggregated mean on fields usage_guest, usage_guest_nice,usage_idle,usage_iowait, usage_steal,usage_system, usage_user
Current version supports only basic HTTP authentication via Simple Login and AWS Cognito. Simple Login method stores username and password in configuration file (in plain text).
Authentication validator module is called dynamically based on the
configuration. Additional modules can be added in the authentication
directory of the project.
-
startServer.pyspins up a werkzeug server usingrun_simplemodule. This is not suitable for production and doesn't perform well on concurrent requests. -
The code can also be hosted AWS Lambda and served with AWS API Gateway for serveless implementation. Code has tested with Zappa framework. Detailed instructions to be released soon.
-
Docker image to be released soon.