Rahul Agarkar created HBASE-29035:
-------------------------------------
Summary: Amount of region cached in the region metrics not updated
for a region immediately after it is flushed with cacheOnWrite turned on
Key: HBASE-29035
URL: https://issues.apache.org/jira/browse/HBASE-29035
Project: HBase
Issue Type: Bug
Components: Balancer, BucketCache
Reporter: Rahul Agarkar
Assignee: Rahul Agarkar
An issue has been observed where if a new region is created and flushed with
the configuration "hbase.rs.cacheblocksonwrite" is set to true, then, the
region metrics does not reflect the amount of region cached for this new
region. It starts reflecting the correct metric when this region is reopened
for whatever reason.
The steps to reproduce this issue are as follows:
* Create a table {{create 'tprefetch', 'cf'}}
* {{{}{}}}Insert a record in the table {{put 'tprefetch', 'row1', 'cf:col1',
'aaaaa'}}
* {{{}{}}}{{{}flush 'tprefetch'{}}}
After the flush, query the region metrics for this region and it will show that
the region is 05 cached. This leads to the cacheAwareLoadBalancer take
incorrect decisions whenever a region server is restarted.
{{}}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
