Bug #21119
Updated by Peter Amstutz 7 months ago
Observed on user cluster: After a few minutes (~15 minutes the first time I observed this, 4 minutes the second time, the third time it went for 39 minutes) I will keep trying and see if there's any pattern) Workbench stops receiving messages from websockets. In the web developer tools, I see a steady stream of messages, which then abruptly comes to a halt. Unlike the previous websockets bug, upon reconnecting I start receiving messages again. Message volume here is pretty high, something like 30 messages per second. Each message is 400-600 bytes. These are mostly container updates. It looks like that, for container updates, we're only publishing "owner_uuid" and "state". In a lot of cases both the old and new states are "Running" so to bring the volume down, we could consider not publishing events where old_attributes and new_attributes are the same.