Looks like the tests might be flapping. I get random failures on the following (with enough travis re-builds it eventually goes green), for example:

--- FAIL: TestNoQuorum-2 (3.08s)
    cluster_test.go:143: Disconnect from wrong server: expected=127.0.0.1:12431, actual=127.0.0.1:12428

2016/04/12 18:29:55 Send loop terminated: err=<nil>
--- FAIL: TestBasicCluster-2 (8.13s)
    cluster_test.go:42: Get failed on node 2: zk: node does not exist

@spenczar where these the kind of errors you were seeing that prompted #76?

I'm not sure, it's been a long time since I wrote that. If I remember right, I wrote that because it was possible for the zookeeper process to start, but not be initialized enough to handle commands. Similarly, shutdown takes a bit of time. That PR just switched from "sleep for 1 second when starting up" to "poll the server every second to see if its up, max 5 seconds," which I think at least can't hurt.

Sorry I can't be more helpful on this specific bug. It's been a long time since I looked at this.

No worries, thanks for chiming in. Without looking at the details in the PR, sounds like a good approach, and could be related to the tests flapping. I'll need to dig in deeper. I hesitate in spending too much time, doesn't look like much active maintenance happening lately.