Polling for Models
The model poller runs periodically to pick up models that are ready for processing.
The XML files for each data model are initially located in the
input
folder. Based on the outcome of the processing, the models are moved to either the
success
bucket or the
failure
bucket and appropriate log messages and notifications are generated.
The model poller process is controlled by the following environment variables:
Environment Variable | Description | Default Value |
---|---|---|
pollerInitialDelay
|
The delay (in milliseconds) after which the first polling iteration starts once the Model Loading Engine micro service starts running. | 60000
(60 seconds) |
pollerFixedDelay
|
The delay (in milliseconds) between consecutive iterations of the polling process. Fixed delay environment variable controls how much gap in seconds needs to be added between two polling events. | 180000
(180 seconds) |
The model poller process is configured such that two iterations of the process start concurrently. The delay between two iterations is controlled by the
pollerFixedDelay
environment variable.
For example, the poller process starts at 10am and the model loading activity takes 10 minutes to complete, and the
pollerFixedDelay
is configured to be 3 minutes. The second iteration of the polling process starts at 10:13am, that is 3 minutes after the first iteration.
The number of threads used for processing models is controlled by the following flag that is configured using
Flag Name | Default Value | Recommended Value | Flag Path |
---|---|---|---|
modelPollerThreadCount
|
100 | 100 threads for each node in the object store cluster.
For example, for a three node cluster, use 300 threads. |
/config/applicaton/tenants/default |