[
https://issues.apache.org/jira/browse/HBASE-10058?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Lars Hofhansl updated HBASE-10058:
----------------------------------
Description:
In HBASE-9915 we found that for encoded blocks the quick check whether the seek
key falls on the current block did not work.
This adds a unittest so we won't regress. The test counts the number of blocks
accessed (real data/index block in 0.94 and cache hits in trunk). Without
HBASE-9915 we see about 50% more blocks hit in this test.
was:
Was just trying to write a unit test to make sure we won't regress on
HBASE-9915 in the future and found the dataBlockReadCnt metric is not correctly
updated for encoded blocks (type is ENCODED_DATA as opposed to DATA).
> Test for HBASE-9915 (avoid reading index blocks)
> ------------------------------------------------
>
> Key: HBASE-10058
> URL: https://issues.apache.org/jira/browse/HBASE-10058
> Project: HBase
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Assignee: Lars Hofhansl
> Fix For: 0.94.15
>
> Attachments: 10058-0.94.txt, 10058-trunk.txt
>
>
> In HBASE-9915 we found that for encoded blocks the quick check whether the
> seek key falls on the current block did not work.
> This adds a unittest so we won't regress. The test counts the number of
> blocks accessed (real data/index block in 0.94 and cache hits in trunk).
> Without HBASE-9915 we see about 50% more blocks hit in this test.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
