Add new features:
1. plugins of Ambari and k8s deploy
2. specified config 'priority_network' to solve some ip problems
Fix bugs:
fix bugs that rebalance does not work in some case.
fix count(*) from union stmt bug
fix some union stmt bugs
fix bugs when try to schema change a clone replica
* add authentication to DownloadAction
1. use cluster_id as token;
2. add dir limit, only files in data dir can be accessed.
* enable authentication in DownloadAction by default
compile failed
* select wrong rollup
* update backend in cluster because of forgeting to remove backend' id when drop backend or decommission in latest version
* aggragation function containing funcion with wrong param crash
* compile failed
* aggragation function containing funcion with wrong param crash
* select wrong rollup
* update backend in cluster because of forgeting to remove backend' id when drop backend or decommission in latest version
* aggragation function containing funcion with wrong param crash
* modify modify the way of PartitionDesc analysis. fix set global variables bug
* add export job statitic proc
* merge from baidu/master
* change MASTER_CLIENT_TIMEOUT
* fix right outer join with null column
* Use epoll instead of select in mongoose.
When number of fds exceed 1024, FD_SET will crush, so switch to epoll
* add default case to epoll event in mongoose read
In case that we want to clone a replica with version X, but a replica with stale version(less than X) is already in the dest backend.
This will leads the clone task to success directly, which is not what we expected.
Adding version check when doing the clone task and try to drop the replica with stale version
