News for hackers. Stuff that matters.
Re:A Review of Airbnb's SRE Hiring Process, from someone who... by MichaelGG on Thursday August 16, 2012 @04:38AM (#4390444)
I was asked how I could coordinate actions between many machines, and I told them that I could probably use Chef to do a lot of that, which was met with a huge red “no” response. I then had to play bingo and landed on the “correct answer” of “Zookeeper.”

I'm curious about this. I'm only vaguely aware of either system, but I thought they solved rather different problems. Chef handles system configuration (like things in /etc, or "configure memcached to listen on this IP and port"). Zookeeper is more of an API to deal with distributed locking/sync issues for an application itself.

Was this interview question less of a "bingo" and more of a "one-side really misunderstood something"? Or how do Zookeeper and Chef compete?

Re:A Review of Airbnb's SRE Hiring Process, from someone who... by cparedes on Thursday August 16, 2012 @04:49AM (#4390480)
Author here. They do, though for a certain set of problems, you can get away with using things like Chef for coordinating lots of information between a lot of machines.

My point, however, was that tools should never be the focal point of solving a problem - if you recognize the problem first as a distributed locking problem, or as a configuration problem, then you can start deriving a solution (and maybe that solution might have Zookeeper in it, maybe you'll write a new piece of software, maybe you'll use Chef, etc.)

Re:A Review of Airbnb's SRE Hiring Process, from someone who... by ryanpers on Thursday August 16, 2012 @05:21AM (#4390557)
One of the biggest problems as I noted in a different comment is that chef doesnt solve _software_ deployment. Configuration management and "best effort file pushes" is not really appropriate for running a website. Sure you could do something with chef, but waiting 20 minutes for a rollback would be pretty lame-o.

The issue here is software push and service orchestration. How do you do a rolling restart of a cluster in an automated fashion?

As for 'configuration management', I feel like puppet, chef, etc arent catch all solutions for 'i want my config to be the same on all machines' for the above mentioned 20 minute cycle.