Cluster panics when duplicate node url is encountered

[cboggs]

Steps to reproduce:

1. Spin up 3 node cluster with Node1 as leader and Node2/3 joining Node1 on initial startup
2. Shut down Node1
3. On Node1, rm -rf /var/opt/influxdb/*
4. Edit Node1 config to join Node2 and Node3 on next startup
5. Start Node1
6. Observe cluster-wide panic
7. Restart InfluxDB on Node2 and Node3 - these two nodes seem to start correctly as long as Node1 stays offline

Here's what I see in the logs on Node2 and Node3:

[raft] 2015/04/14 15:30:04 apply: add node: duplicate node url
panic: apply: add node: duplicate node url

goroutine 8 [running]:
log.(*Logger).Panicf(0xc20800a730, 0x9ed850, 0x13, 0xc2083f3d58, 0x1, 0x1)
    /root/.gvm/gos/go1.4.2/src/log/log.go:200 +0xd1
github.com/influxdb/influxdb/raft.(*Log).mustApplyAddPeer(0xc20803a820, 0xc208392c90)
    /root/.gvm/pkgsets/go1.4.2/global/src/github.com/influxdb/influxdb/raft/log.go:1518 +0x2ce
github.com/influxdb/influxdb/raft.(*Log).applyNextUnappliedEntry(0xc20803a820, 0xc20800c6c0, 0x0, 0x0)
    /root/.gvm/pkgsets/go1.4.2/global/src/github.com/influxdb/influxdb/raft/log.go:1436 +0x752
github.com/influxdb/influxdb/raft.(*Log).applier(0xc20803a820, 0xc20800c6c0)
    /root/.gvm/pkgsets/go1.4.2/global/src/github.com/influxdb/influxdb/raft/log.go:1382 +0x161
created by github.com/influxdb/influxdb/raft.func·002
    /root/.gvm/pkgsets/go1.4.2/global/src/github.com/influxdb/influxdb/raft/log.go:389 +0x764

[cboggs]

Deleting the 'failed' node via the API succeeds, and the dead server no longer shows up in "show servers" queries, but trying to start that node again causes the panic all the same.

However, deleting the server via the API before breaking it (by deleting the data dirs) actually allows the node to rejoin the cluster. Awesome!

Just need to iron out the "panic when a node dies before it is explicitly removed from the cluster" behavior, I think.

[jwilder]

Related #1471 #1472

[otoolep]

No longer applicable in new design.