Hello. One of our Customers want to know about some features in our roadmap.
Anybody know will be such funcional or not:
Manual AI training. For example we have problem with possible root-cause - 3 servers. Operator chooses one of the three servers after internal analisys as main root cause of the problem.
So next time AI will show this server/process/service as main root cause of the problem.
If you can provide me some links, please, share with me.
This is a tricky situation. It seems to me that if one of the server is the root cause of a problem, there is no guarantee that it will be the root-cause of the next problem(s).
You can try to leverage the power of Davis AI as Wolfgang refers to in the URL below, but I believe that's not a correct solution to your problems.
I've heard on several occasions on how Dynatrace will open up how we can interact with Davis, even how it could be "teached" into learning things, but I still think it's not available yet.
Davis always works based on the ranking of potential issues based on events. In some rare cases, typically when tracing data is incomplete, you might end up with "incorrect root cause". First I'd recommend validating if tracing through your apps works correctly. Additionally, put additional metrics when necessary (connection pools for example) and events to "feed" the AI with data.
Davis is marketed as a "deterministic" or "explainable" AI. It will always have the same conclusions on equal inputs. But you cannot tune it by yourself, you can only tune anomaly detection for your entities and add additional metrics or event to it.