Heya guys, we're looking to roll out Growthbook op...
# ask-questions
e
Heya guys, we're looking to roll out Growthbook open source ontop of a snowflake instance. Is there a way to use public private key authentication with growthbook or does it have to be username + password?
f
Hi Moss. I don't think you can yet. Is this a blocker for you?
e
Hmm, it's a bit of a pain as we tend to use pub priv for all system to snowflake connections. I'm having a squiz to see if it's supported in snowflake-promise the library it looks like you guys are wrapping.
Cheers for the quick response @fresh-football-47124! We'll see how we go, we might play with a branch and put in a PR if we get it to work. We're planing on seeing if we can deploy growthbook.io in a hackthon Thursday / Friday this week so you might here from us a bit in this channel over the next couple of days ๐Ÿ˜„
f
okay
Love a good hackathon - please let us know how we can help
e
Cheers man, will do! ๐Ÿ™‚
Are there any things worth considering that aren't in the readme that we should pause and give some deeper consideration to if we're setting it up quickly? The main one I can see is making sure we find a suitable balance between how much of the metric definition happens in the warehouse +dbt vs in the metric statements. Leaning towards most of the metric processing happening in wh to allow for relatively clean queries in the metrics layer.
f
ya, it's a bit of a trade off. I would tend to agree with your approach. The experiment queries do several joins which can get ugly if the metrics themselves have lots of joins.
๐Ÿ‘ 1
g
For what its worth @early-battery-30877, weโ€™re doing the latter: so we create separate tables per metric in the warehouse specifically for growthbook (but also because we have tests on them there). Still trying to figure out if this is a good approach though.. @fresh-football-47124 not sure if this would be an option but is there any particular reason the group assignment and the metrics are separate sets that are joined? I can see both pros and cons here (the con being, in my case they are both a few billion rows big).
f
hi Robin - we might be able to make it work better for your case, would you be up for a meeting so we can understand your use case?