etcdserver/api/etcdhttp: add reason field for /health response

[tangcong]

Fix issue #11599 . @gyuho

{"health":"false","reason":"NOSPACE"}

[gyuho]

@wenjiaswe @jingyih Can we make sure kube-apiserver /healthz be compatible with this change?

[tangcong]

it looks compatible with kube-apiserver /healthz.

https://github.com/kubernetes/kubernetes/blob/2446c3d7e96008c1da59d79e1518659028b04267/staging/src/k8s.io/apiserver/pkg/storage/etcd3/healthcheck.go#L24-L40

[codecov-commenter]

Codecov Report

Merging #11983 into master will increase coverage by 0.39%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master   #11983      +/-   ##
==========================================
+ Coverage   65.50%   65.90%   +0.39%     
==========================================
  Files         403      403              
  Lines       37320    37323       +3     
==========================================
+ Hits        24447    24598     +151     
+ Misses      11337    11176     -161     
- Partials     1536     1549      +13     

Impacted Files	Coverage Δ
etcdserver/api/etcdhttp/metrics.go	41.37% <0.00%> (-7.61%)	⬇️
client/keys.go	68.34% <0.00%> (-23.12%)	⬇️
auth/simple_token.go	74.78% <0.00%> (-14.29%)	⬇️
auth/jwt.go	64.04% <0.00%> (-13.49%)	⬇️
clientv3/naming/grpc.go	63.15% <0.00%> (-12.29%)	⬇️
etcdserver/api/v3rpc/lease.go	72.15% <0.00%> (-7.60%)	⬇️
proxy/grpcproxy/watcher.go	85.71% <0.00%> (-4.09%)	⬇️
clientv3/leasing/cache.go	87.22% <0.00%> (-3.89%)	⬇️
proxy/grpcproxy/watch.go	89.94% <0.00%> (-2.96%)	⬇️
lease/leasehttp/http.go	61.31% <0.00%> (-2.92%)	⬇️
... and 22 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 3082a7d...3304241. Read the comment docs.

[wenjiaswe]

Wenjia Jingyi Hu Can we make sure kube-apiserver /healthz be compatible with this change?

k8s liveness probe and storage backend health check should be fine, it's not decoding the body but use the http code to verify.

[tangcong]

Wenjia Jingyi Hu Can we make sure kube-apiserver /healthz be compatible with this change?

k8s liveness probe and storage backend health check should be fine, it's not decoding the body but use the http code to verify.

thanks. it is better if we can print a clear error message in apiserver health check. @wenjiaswe @xiang90 @gyuho

[tangcong]

issue #11599 describes /health still returns {"health": "true"} when etcd has exceeded DB and detected DB corruption.It looks like the return is correct {"health": "false"} in master and release-3.4 branch. do you have any suggestions for this pr ? thanks. @gyuho

[gyuho]

lgtm thanks @tangcong @wenjiaswe