SIMCA®-online hangs when Batch Context Generator is populated with large amounts of data (Q938)
Symptoms:
The SIMCA®-online server and all clients hang and become unresponsive. This can happen when a server is started and it needs to load a large amount of data using the Batch Context Generator.
Long delays and client unresponsiveness can also occur when 'Find batches' is used for a long time span for a Batch Context Generator node that needs large amounts of data. In this case the server does not hang, but all clients have to wait until Batch Context Generator is done requesting data.
Cause:
The server unnecessarily locks an internal database when Batch Context Generator fetches data, which can cause a temporary deadlock in the server and all clients become unresponsive.
Resolution:
A 16.1.1 patch has been released to address this issue. See SIMCA®-online 16.1.1.
- The SIMCA®-online client that does 'Find batches' still has to wait until the Batch Context Generator has read the selected time range. When this is happening the client window can become black but will return to normal when 'Find batches' finished.
- The server no longer hangs during startup.
- Other clients that wants to use the same SimApi connection to the data source, for instance doing 'Find batches' for another configuration or extracting data will experience slowness but this is due to using the same connection. Other operations such as logging in, viewing plots etc. should work as expected.
Note: Batch Context Generator still takes a long time to populate data after the fixed version has been installed, and this still affects the entire server. For example, some actions in Project Explorer will still wait until the Batch Context Generator node is finished populating: changing a project configuration, starting and stopping execution, and prediction of finished batches.
Status:
Fixed in SIMCA®-online 16.1.1.
More Information:
It is recommended to configure Batch Context Generator nodes using the recommendations from the help, so that they don't become too slow (for example using a too short sampling interval, or configuring the batch length incorrectly).
Also, if you have batch contexts in your data source, such as OSIsoft PI Event Frames, the recommendation is to use that because then the data source becomes the single source of truth.
