How to using HBaseWD is better.

[Paul Yang]

HI!

We imported the hbasewd lib into our online project.

I have a feeling there is an issue in our project.

there are our environment:

HBase 5 region servers

hbasewd-0.1.0

Hadoop CDH3u5

HBase ROW KEY: hashPrefix + timestampe + MD5(business key)

50G writing to HBaseWD table per day

We did it with RowKeyDistributorByHashPrefix.class and there are 120 buckets. Pre-splitting table with 120 buckets when writing to it with HBaseWD.

So

1. There are just 120 activity regions online, Other regions will can't be written, Because they are out of the rowkey range. It will increase regions split frequency. Maybe there is a bottleneck of writing preference. Yes or No?

2. How to avoid the issue, should we add the buckets to 240? or split the data to other hbase table.

3. Usually, how many regions for writing per one region server is normally?

Thanks for your help!

Paul