[
https://issues.apache.org/jira/browse/HBASE-29326?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Wellington Chevreuil updated HBASE-29326:
-----------------------------------------
Fix Version/s: 2.7.0
> Do not prevent prefetch from running if the given file has already been fully
> cached before
> --------------------------------------------------------------------------------------------
>
> Key: HBASE-29326
> URL: https://issues.apache.org/jira/browse/HBASE-29326
> Project: HBase
> Issue Type: Bug
> Affects Versions: 3.0.0-beta-1, 4.0.0-alpha-1, 2.7.0
> Reporter: Wellington Chevreuil
> Assignee: Wellington Chevreuil
> Priority: Major
> Labels: pull-request-available
> Fix For: 3.0.0, 4.0.0-alpha-1, 2.7.0
>
>
> Since HBASE-28176, we were skipping prefetch on files that have been fully
> cached before (and where BucketCache persistent cache successfully recovered
> all data blocks in the event of restarts). The problem is that meta blocks
> would be skipped too, and would only make into cache upon client reads, which
> could be costly when using cloud storage. In the prefetch logic, we already
> try to read each block from the cache first and skip file reading if we find
> the block already present in the cache. That would cause minimal impacts for
> running prefetch again on recovered bucket caches, as only meta blocks would
> indeed require cloud storage access.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
