siridb / siridb-enodo-hub Goto Github PK

• Filter by if graph is stationary.
  -..

add split and regex search/match to jinja allowed helper functions in template

• Create a process with a pipe connection to SiriDB to receive incoming data

Add configuration for:

• default min period for data to analyse for a pattern
• Siridb server credentials, to query for base/history data

TLDR;

We need to make a realtime worker for anomaly detection

Description

In order to process series fast enough we need a worker that holds an forecast result for a series as state. When a new datapoint comes in, it should be evaluated within a very short timespan.

Features

• Request forecast via Hub from other worker
• Hold datapoint count to check if it is large enough
• Hold errors from forecast result if any

filter fields in template get request

Is there any tourial for enodo quick start?

• Create event state / history
• Create auto close template
• Collect response from external system to be used in auto close template

TLDR;

We need a new client to the worker, which handles the incoming requests, and send result or job requests to the Hub

Description

To accomodate the new worker, we can update the worker client in the lib

Features

• Worker holds state
• Worker cleans state of series after N-amount of time
• Worker can be queried for data
• On a friendly shutdown the worker should save its state to disk. This will help reduce the amount of siridb queries it has to make on startup

TLDR;

We need to remove the scheduling part from the Hub into a seperate service. the remaining functionality of the hub will stay in the Hub service, with the difference that multiple Hub instances are possible, and will be kept up-to-date via ThingsDB

Description

The Hub service handles API requests and incoming datapoints that the listeners send through. The Hub services holds a list with series names and their config.

Functionality

Some agreements on functionality:

• We explicit insert series for realtime analysis. This will be a property on the series config
• Hub has no state only config
• Bulk endoints for inserting and querying
• Hub sends worker config to worker on handshake
• Hub should be able to restart without any complications. All data should go via thingdb
• The Hub uses an algorithm to determine which series goes to which worker, for easy scalability of the workers