MisterWang created HBASE-28783:
----------------------------------
Summary: Concurrent execution of normalizer operations on tables
in RegionNormalizerWorker
Key: HBASE-28783
URL: https://issues.apache.org/jira/browse/HBASE-28783
Project: HBase
Issue Type: Improvement
Components: Normalizer
Reporter: MisterWang
Recently, I have been managing the large tables in the HBase cluster by
enabling the normalizer to set the size of the regions and keep the number of
regions within a reasonable range.
The current code retrieves tables from RegionNormalizerWorkQueue and performs
normalization operations on the tables in series. When there are multiple large
tables in a cluster that need to be managed, the efficiency will be very low.
I have confirmed that each split or merge plan generated by each table during
the normalizer process will be limited by RateLimiter, so I think it is
reasonable to perform table normalizers concurrently.
In terms of implementation, create a thread pool for executing tasks in
RegionNormalizerWorker, with a default value of 1 for the number of thread
pools, and provide a parameter that can be configured to other values.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
