Indexing Pipeline Extension State
Indexing Pipeline Extension State
An IPE may need to perform web requests to access an external resource required for its execution, such as an authorization token or a catalog.
In order not to flood the remote server, you can keep and share an IPE state between executions using the state
keyword.
The state
must be a Python dictionary and has a 32kb size limit.
Oversized state or invalid types will be dropped.
The state is shared between the instances of a single IPE bound to a single source.
For example, a state stored for source_1
cannot be accessed from source_2
, and neither can a state stored for IPE_1
be accessed from IPE_2
.
The state is cleared every 24 hours. You must implement your own logic in order to support a shorter lifetime. Since IPEs are executed in parallel, it may take up to a minute to propagate the updates to all running instances of the IPE.
token = state.get('token')
if token:
# Use existing token in request
# ...
else:
# Set the token in the state, and move on
state['token'] = 'new_token’
# ...