Refine etcdctl snapshot restore error logs

[hongchaodeng]

Continue discussion: #9179

The following lines are always printed to stderr

2018-01-19 23:42:54.398661 I | pkg/netutil: resolving test-etcd-backup-restore-0-0000.test-etcd-backup-restore-0.e2e-etcd-operator-flaketest-580.svc:2380 to 10.28.5.36:2380
2018-01-19 23:42:54.402454 I | pkg/netutil: resolving test-etcd-backup-restore-0-0000.test-etcd-backup-restore-0.e2e-etcd-operator-flaketest-580.svc:2380 to 10.28.4.183:2380

It has two issues:

• It causes confusion. These seems normal message. As a user I wouldn't expect to see it on stderr for successful run.
• For DNS resolving issues, the DNS and IP were already included in the error message: *: add detailed error message on URLs equal check #9210. So the messages are duplicate.

[xiang90]

It causes confusion. These seems normal message. As a user I wouldn't expect to see it on stderr for successful run.

No. This is the whole point of why stderr exists. Users should only redirect stdout if they only care about the final result rather than any informational or diagnostic messages.

[hongchaodeng]

@xiang90 I understand your argument. But let's be honest that this mixing of diagnostic information and user-facing error message is confusing.

If etcdctl runs successfully, it logs what it affected users like writing files. If it fails, it logs what failures should user expect -- in this case, Error: --initial-cluster ....
Isn't it better to separate diagnostic and user messages?

As a side note, stdout/stderr doesn't really matter.

[gyuho]

I've improved error messaging to be more clear. And now you can specify log output (default is stderr) with --log-outputs stdout,a.log flag.