Filip Petkovski

Yes, I can try this again next week, currently I am OOO. The drop seems to be exactly 50%, so I was wondering if we were either doing something more...

I tried the lazy approach and the `grpc_client_handled_total` metric was almost the same as in the eager approach. However, I did feel like the individual queries were faster with the...

Sorry actually in the lazy implementation I do see an issue. In the logs I get see: ``` Internal desc = runtime error: invalid memory address or nil pointer dereference"...

Ha, the e2e test failure seems to be exactly what I saw :)

This is a limitation with the Prometheus TSDB. The only way to backfill data older than a few hours is to upload TSDB blocks directly to S3 from where you...

@matej-g would you mind rebasing this PR? I'd like to do another review.

I would say this is rather a Kubernetes, or more precisely, a kubectl problem. Even if we shrink the CRD now, at one point it might grow again. @sathieu if...

Still relevant.

Memory snapshotting seems to be a TSDB option. Maybe it's just a matter of setting this flag when starting receiver TSDBs?

A lot of mount paths are implementation details and not relevant for the user. Could you explain what you're trying to accomplish?