GitHub user tenthe added a comment to the discussion: Monitoring Adapters
Hi @kruscho, I fully agree, extending the metrics you described makes a lot of sense. However, calculating throughput for all adapters might be challenging. For broker-based adapters this is feasible, but for pull-based adapters it’s unclear how throughput should be defined, since we don’t have actual incoming events. Maybe we can define a sucessfull read (`pull`) as an incoming event? I assume the core idea behind this feature is to detect when a data source stops providing data. That would indeed be valuable, but I’m not yet sure what the best way to model this would be. One option might be to introduce a “data source health” signal, but this would need to be implemented individually for each adapter. Another open question is how to define abnormal behavior. For example, machines being turned off is perfectly normal, and users typically don’t want to receive alerts in those cases. However, if a machine is turned on and the OPC UA server, for instance, stops sending data, then we do want to trigger a notification. Overall, I'm still not sure what the best solution might look like. These are just a few initial thoughts. What do you think? GitHub link: https://github.com/apache/streampipes/discussions/3936#discussioncomment-15058632 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
