Are we comfortable to upgrade etcd server to a new version now?

The risk of this falling over is scary - what's the reward?

Very scary, esp since I have no way of telling what source was sitting on your computer when you pushed the button.

Very scary, esp since I have no way of telling what source was sitting on your computer when you pushed the button.

Not disagreeing, but can you explain? Isn't the makefile just building a Docker container from a binary it fetches? How is source code sitting on his computer related?

The makefile builds source.

For issue #10122, @cjcullen and I discussed, the current proposed solution is not starting any pods / containers until network configuration is done. In this case, shouldn't no rush to upgrade to a new etcd version?

etcd 2.0.9 was also built from someone's desktop. After chatting with a few people and looking at the CLs pulled into 2.0.12 -- especially fixes for bugs that cause an etcd install to become semi-permanently unhealthy -- it seems like a reasonable approach to switch to it as the least-worst option.

And if the release-of-etcd process is scary, we should figure out a sustainable long term approach but not block this particular PR on it.

LGTM based on our offline discussion.

I am going to merge this one for now, so that @zmerlynn can cut 0.19.3 to include this one. @zmerlynn please make sure you cut 0.19.3 is solely based on 0.19.1 + this pr. Thanks!

On it.

... and apologies to those not involved in offline discussions.

Note that if we find issues with 2.0.12 in the RC validation process, we can roll it back.

Ok to test.

I'm going to go ahead and merge this - we've had a lot of internal soak time with these bits from a private build.