We shouldn't try to take two snapshots at the same time

[jvshahid]

As mentioned in this thread, if we tried to take two snapshots at the same time, the second snapshot will error out with Last snapshot is not finished. TakeSnapshot() in raft_server.go should lock to make sure that we don't try to take multiple snapshots at the same time. Note that taking two snapshots in a row doesn't necessarily do extra work, since raft will ignore the command if no commands were committed since the old snapshot.

[consense]

+1

[dgnorton]

Can be reproduced using https://github.com/dgnorton/influxdb_stress.

• create a queries.txt file with just one query in it drop series {{randSeries}}
• run ./influxdb_stress -q queriest.txt

[toddboom]

@dgnorton This might be related to the issue I saw last night. It seems that lots of drop series under heavy read/write load was able to get a cluster into a wonky state.

[dgnorton]

@toddboom investigating now to see if more than one server from raft_server.go is being created.