James Blackwell

We use FPM to build Debs of carbon-c-relay for the particular distro and version that we use. This does not have anywhere near all of the pieces that you need...

no-service-when-index-is-not-ready might actually work! I was concerned that carbonserver would pass a 5xx up to graphite-web, which would break all the queries with grafana, but that seems to not be...

> I'm not seeing any crazy queries at the moment, this looks like we get it pretty rough at times. I suspect things go sideways when we get a series...

Ahh, I set that to a million based on the commented out request. What's a sane ceiling that we can set? 100k? 10k? ``` # Maximum metrics could be returned...

In regards to compression, we might be do that now, now that we've gotten graphite-web kicked off of the caches. I'm supposing that turning compression on (or off) will cause...

This seems more user error and less bug. I'd remove the bug label and replace it with question if I could, but it seems I lack the ability to do...

Nope. The only thing that reads and writes whisper these days is go-carbon. Our caches in this cluster hold close to a terabyte of data each, making the movement of...

> > The thing that's really throwing me here is this is one of a cluster of 50 with rf=1. A cluster that as recently as last week with 64...

This issue started causing us trouble when we moved our cache layer to carbonserver. I have created a PR https://github.com/go-graphite/go-carbon/pull/453 that optionally makes an empty result valid.

Thank you for the very quick response, @deniszh! I have added trie-index = true and restarted go-carbon. Unfortunately I'm still seeing a substantial delay in the visibility of created metrics....