You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@skywalking.apache.org by Sheng Wu <wu...@gmail.com> on 2020/08/06 07:12:08 UTC
Re: Es storage balance
Please follow this for more details,
https://github.com/apache/skywalking/issues/5251
Sheng Wu 吴晟
Twitter, wusheng1108
刘 嘉鹏 <Ev...@outlook.com> 于2020年8月6日周四 下午2:38写道:
> Hi,all:
>
> Nowadays,skywalking index rolling step is same in metrics data and record
> data.<http://data.in>
> in<http://data.in> default es cluster shards balance strategy is balance
> es shards number on every node.
> So in some case:
>
> one node have 20 metrics shards and 10 records shards. another has 20
> record shards and 10 Metrics shards.
> at cluster level. all nodes is balanced.
> but in node level. not balance
>
>
> In our cluster ,some case is like following:
>
> https://user-images.githubusercontent.com/31562192/89490048-6d9def80-d7de-11ea-9c99-90a574af9abc.png
>
> So we want to enhance the feature:
> 1. Add a field called dayStepForRecord and the original dayStep is for
> metrics data
> In production: we should set record daystep is 1,and the metrics day step
> is 15 or more.
>
> Then the ElasticSearch LB mechanism should be back to work again
> Evan.liu
> Thx for your reply
>