pd leader balance radio 100% and region balance ratio 100%

[laizhebujum]

请大神答疑，为了更好的描述问题直接中文啦~

overview页和pd页都显示平衡比例100%（图1），但是查看tikv监控已经达到平衡（图2）

通过pd-ctl查看tikv store的leader和region信息也是正确的，已经达到平衡。（图3）

说一下最近做过的操作：
1、2018-12-10 tidb版本是2.0.8，本想做滚动升级2.1，却遇到点小插曲：
升级过程中报错FAILED! => {"changed": false, "elapsed": 300, "msg": "Timeout when waiting for search string 200 OK in xxxxxxxxx:2379"}，以为是这个pd有问题，就缩容去掉了这个pd，后来发现是inventory.ini deploy_dir参数配置有误，然后又添加上这个pd了，（添加删除启动过程都是用的2.0.8版本的tidb-ansible），集群启动正常，版本是2.0.8

2、2018-12-11日 参照官方文档做了滚动升级至2.1，升级成功了，集群正常启动 下载tidb-ansible日期是2018-12-11，就出现了balance radio 100%的问题了。

[disksing]

Hi @laizhebujum, does your cluster have tombstone store(s)? It looks the balance ratio could be incorrect when there are tombstone store(s).

[laizhebujum]

Hi @laizhebujum, does your cluster have tombstone store(s)? It looks the balance ratio could be incorrect when there are tombstone store(s).

Thank you for your reply.
Yes, it does exist. It already exists in version 2.0.8, but there was no problem.
How can we clean up the tombstone store(s)？
In addition, after upgrading to 2.1.0，There is such information in the log every minute：
2018/12/13 17:27:29.933 cluster.go:472: [info] buried store id:1 address:"X.X.X.X:20160" state:Tombstone

[disksing]

Because the statistical adjustments made the previous calculations unsuitable, we are planning to replace the original ones with better calculation formulas. At the moment you may have to choose to ignore the balance ratios, sorry about that.
The buried store log message should be a bug (disturbing but not critical). Thanks for your report!
/cc @rleungx

[laizhebujum]

Thanks

[rleungx]

The disturbing log has been removed in #1382.

[rleungx]

We remove this metrics in release 3.0 and provide an API to clean the tombstone store in release 2.1 via #1705. Closes this for now.