etcdctl: role add 'empty' error

[guanwenbogit]

Please read https://github.com/etcd-io/etcd/blob/master/Documentation/reporting_bugs.md.
when I exec this
./etcdctl role add ''
Well, the role name is empty string .Then I got this:
Error: rpc error: code = Unavailable desc = transport is closing

Even, the etcd server crashed, and it can not be started any more !!!
The etcd server log :
2019-07-18 21:53:21.843673 I | etcdmain: etcd Version: 3.3.0+git 2019-07-18 21:53:21.843723 I | etcdmain: Git SHA: 948e276ca 2019-07-18 21:53:21.843731 I | etcdmain: Go Version: go1.12.1 2019-07-18 21:53:21.843736 I | etcdmain: Go OS/Arch: darwin/amd64 2019-07-18 21:53:21.843745 I | etcdmain: setting maximum number of CPUs to 4, total number of available CPUs is 4 2019-07-18 21:53:21.843755 N | etcdmain: failed to detect default host (default host not supported on darwin_amd64) 2019-07-18 21:53:21.843760 W | etcdmain: no data-dir provided, using default data-dir ./default.etcd 2019-07-18 21:53:21.844305 N | etcdmain: the server is already initialized as member before, starting as etcd member... 2019-07-18 21:53:21.900200 I | embed: name = default 2019-07-18 21:53:21.900223 I | embed: data dir = default.etcd 2019-07-18 21:53:21.900230 I | embed: member dir = default.etcd/member 2019-07-18 21:53:21.900237 I | embed: heartbeat = 100ms 2019-07-18 21:53:21.900243 I | embed: election = 1000ms 2019-07-18 21:53:21.900249 I | embed: snapshot count = 100000 2019-07-18 21:53:21.900267 I | embed: advertise client URLs = http://localhost:2379 2019-07-18 21:53:21.900277 I | embed: initial advertise peer URLs = http://localhost:2380 2019-07-18 21:53:21.900286 I | embed: initial cluster = 2019-07-18 21:53:21.953318 I | etcdserver: restarting member 8e9e05c52164694d in cluster cdf818194e3a8c32 at commit index 8 {"level":"info","ts":"2019-07-18T21:53:21.959+0800","caller":"raft/raft.go:661","msg":"8e9e05c52164694d became follower at term 3"} {"level":"info","ts":"2019-07-18T21:53:21.960+0800","caller":"raft/raft.go:383","msg":"newRaft 8e9e05c52164694d [peers: [], term: 3, commit: 8, applied: 0, lastindex: 8, lastterm: 3]"} 2019-07-18 21:53:21.996324 W | auth: simple token is not cryptographically signed 2019-07-18 21:53:22.015671 I | etcdserver: starting server... [version: 3.3.0+git, cluster version: to_be_decided] 2019-07-18 21:53:22.016860 E | etcdserver: cannot monitor file descriptor usage (cannot get FDUsage on darwin) 2019-07-18 21:53:22.099921 I | etcdserver/membership: added member 8e9e05c52164694d [http://localhost:2380] to cluster cdf818194e3a8c32 2019-07-18 21:53:22.105184 N | etcdserver/membership: set the initial cluster version to 3.3 2019-07-18 21:53:22.105257 I | etcdserver/api: enabled capabilities for version 3.3 2019-07-18 21:53:22.108232 I | embed: listening for peers on 127.0.0.1:2380 2019-07-18 21:53:22.109467 C | mvcc/backend: cannot put key into bucket (key required)

However, after removing the "default.etcd/member" member-dir, the etcd server can start.
More seriously，if these happened in the etcd cluster, all the members wouldn't start any more.
Unless remove the member-dir one node by one node.

etcd Version: 3.3.0+git
Git SHA: 948e276
Go Version: go1.12.1
Go OS/Arch: darwin/amd64

[spzala]

The key required error is from bbolt. We probably should not allow empty role similar to empty user or member as a quick fix /cc @xiang90 @jingyih (I will create a PR if that makes sense) Thanks!

[jingyih]

Fixed by #10907.