A little elaboration about on [min/max]-block-duration documentation.

[ivan-kiselev]

MR grew out from this thread https://improbable-eng.slack.com/archives/CA4UWKEEN/p1553679266828300.

TLDR: end users, like me, get confused by prometheus_tsdb_compactions_total being incremented and Prometheus logs about compaction when internal Prometheus compaction is supposed to be disabled, so there's a little elaboration for that.

I haven't found a better place in the repo to put this piece of information, tell me if there is such.

[GiedriusS]

Perhaps we should mention that with #838 being merged we will check this automatically? So it's not just a recommendation, but a requirement. What do you think?

[ivan-kiselev]

This PR is definitely worthy to mention since it has a direct relation to the topic, but I'd argue with "a requirement" statement. Running Thanos with sidecar.validate-prom is optional and up to end user so as running compactor itself. There might be quite successful set-up cases without compactor at all, and AFAIK (correct me if I'm wrong) - things will be working with any initial block size in a relatively successful way with obvious downside when the set-up has a really old and huge bunch of data.

Should I make a PR to PR with corresponding changes?

[bwplotka]

Just put those changes in this very PR @homelessnessbo (: It's small enough.

[ivan-kiselev]

👍 Will include it in #838 in a few minutes

[ivan-kiselev]

https://github.com/yeya24/thanos/pull/1/files