some time delay (0.3s) for swebhdfs + kerberos + observer setting.

Settings:

1 master namenode (A), 1 standby namenode (B), 1 observer namenode (C).

following 

https://hadoop.apache.org/docs/stable/hadoop-project-dist/hadoop-hdfs/ObserverNameNode.html

except that 

dfs.client.failover.observer.auto-msync-period.<nameservice>

is set to -1 (not auto -msync)

uable to do curl - - negotiate -u ':' 'https://<observer>:<port>/webhdfs/v1/...'

because it seems like due to the following issue:

https://issues.apache.org/jira/browse/HDFS-14443

using curl --negotiate -u ':' 'https://<master>:<port>/webhdfs/v1/...'

can successfully get 307 redirect with the corresponding Location.

but got 

token (token for xxx HDFS_DELEGATION_TOKEN owner=xxx renewer=xxx masterKeyID=ooo) can't be found in cache"

if redirect the url within 300ms.

Not issue if waiting for more than 300ms and then do the redirect.

No issue if changing (C) to Standby (no observers) (and redirect within 10 ms)