Currently all the logs from everywhere is written into one big log stdout.
It's hard to figure out how is each diagnosis is doing for individual app, and even harder to see when an API failure happened as it's burried in this big log.
With more and more features we're adding, I think we should have a strategy how we improve our logging so we have more visibility what's going on.
My initial suggestion is to only log to stdout API errors, and all analyzer data science work we can create a separate log per job, and have a internal UI like what we've done for deployer to show each job's specific log.
Currently all the logs from everywhere is written into one big log stdout.
It's hard to figure out how is each diagnosis is doing for individual app, and even harder to see when an API failure happened as it's burried in this big log.
With more and more features we're adding, I think we should have a strategy how we improve our logging so we have more visibility what's going on.
My initial suggestion is to only log to stdout API errors, and all analyzer data science work we can create a separate log per job, and have a internal UI like what we've done for deployer to show each job's specific log.