Metrics setup at Robinhood.

[Aravind Gottipati]

Hi,

We use OpenTSDB for our metric collection and wrote about our setup here.

https://medium.com/robinhood-engineering/metrics-collection-and-viewing-7d8a8745dd6

I realize that is old hat to most of you here, but I figured someone might find some use for
the helper scripts etc, linked in the article.

I am happy to answer any questions you might have about this setup, etc.

Aravind.

[Jonathan Creasy]

Thank you, this is great!

[ManOLamancha]

Thanks a ton for the great writeup! What kind of throughput are you seeing? also are you hosting HBase on AWS? 

[Aravind Gottipati]

On Tue, Apr 25, 2017 at 1:49 PM, ManOLamancha <clars...@gmail.com> wrote:
>
> Thanks a ton for the great writeup! What kind of throughput are you seeing?
> also are you hosting HBase on AWS?

Cool, glad you liked it.

We push about 20k metrics per second (at peak) into each of our
OpenTSDB instances. For now, we are not running this on a (HBase)
cluster. We run multiple OpenTSDB instances (with a local HBase) that
get the same data, and loadbalance Grafana across them for
query/visualization performance. Since we push the metrics into Kafka
first, it's relatively easy to spin up multiple OpenTSDB instances
that independetly consume the same metrics stream.

We tried the hosted AWS HBase solution, but we decided against it (for
performance/durability concerns).

Aravind.