EvanLjp edited a comment on pull request #5914: URL: https://github.com/apache/skywalking/pull/5914#issuecomment-735548071
> I think this feature is good for me. But there are some questions in my mind: > > > > 1. Only collect log when having an active span. But we all know the active span depends on the coverage of agents. Maybe we would lose some usage logs when no agent coverage. > > 2. When we coupling the logs and trace in the segment, maybe this way would add more memory cost. For example, in our department, the maximum throughput is about 1000 ops in a pod, and the average size of each log is about 7.5KB. I am not sure about the final cost when using the feature in production. > > 3. Extract metrics for Log to observe or monitor is a common way of production. This kind of sending adds complexity. > > @wu-sheng I prefer to add more appenders on each log framework to collect log. The user often uses TraceId or SpanId to search log on Trace Query UI. But we also support an original way to query, such as the slogan ' just grep' in Loki. Sorry，the above comment ignores the premise of only collecting errors. If only collect the error log, I think it can be added as an optional plugin. The above comment will not be deleted, and can be used as a reference for us when adding the log collection in Satellite. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
