Whats the validity of the logs when they can be boiled down to metrics instead way more concisely (at least with olap-like storage). For traces, as mentioned e

Whats the validity of the logs when they can be boiled down to metrics instead way more concisely (at least with olap-like storage). For traces, as mentioned elsewhere sampling is key. Some systems use % of requests, and other systems cap traces to traces/sec. Logs are just expensive but easy to use, so the best strategy is to lean on filtering to avoid a bunch of pointless repetition which don't add value. TLDR, metrics require more planning but can end up being much cheaper than logs.