DIY Jepsen Testing CockroachDB

cockroachlabs.com

83 points·state_machine··8 comments

This is a great write-up. Easily one of the most interesting posts I've seen on this subject, since Aphyr's last post.

Also nice to see things like:

> We noticed that the Jepsen test code didn’t actually check for negative balances, contradicting what has been previously written about it, so we fixed that.

This is refreshing to see. Much more refreshing than vendors trying to move the goal posts when Kyle's test show them coming up short.

Excellent writeup. Looking forward to more development on this. Hopefully Kyle has time to do a Jepsen writeup (orrrr you can hire him!)

Employee here - A proper analysis from Kyle is certainly on our radar, although we'll likely wait until post-beta since its such a hefty project. We're just as eager to see how CockroachDB stacks up.

makes sense!

These guys should pony up some consulting dollars to get the main guy to do his thing.

Also, I don't buy "NewSQL" without honest CAP discussions of tradeoffs.

Employee here - which discussion around NewSQL would you like to have? You pay for the consistency in more round trips between replicas, and you need a quorum of your nodes (which carry your data) to be alive. All of that stuff can be made arbitrarily more complex to increase throughput, but essentially these are the tradeoffs.

Re: the consulting dollars. Obviously, this blog post had not published without prior contact with Aphyr.

Reply on news.ycombinator.com