You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Deep Golani (Jira)" <ji...@apache.org> on 2025/10/22 16:28:00 UTC
[jira] [Comment Edited] (HBASE-29568) Allow for a configurable "grace period" when using Time Based Priority
[ https://issues.apache.org/jira/browse/HBASE-29568?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18032169#comment-18032169 ]
Deep Golani edited comment on HBASE-29568 at 10/22/25 4:27 PM:
---------------------------------------------------------------
Hi [~wchevreuil],
I was going through the codebase, and from my understanding, we need to introduce a new config key like {{hstore.datatiering.grace.period.millis}} in {{{}DataTieringManager.java{}}}, and modify the {{shouldCacheFile}} logic in {{BucketCache.java}} to respect this grace period for CFs marked accordingly.
Does this approach align with what you had in mind, or am I missing any other part of the flow where the grace period should be applied.
was (Author: JIRAUSER309935):
Hi Wellington,
I was going through the codebase, and from my understanding, we need to introduce a new config key like {{hstore.datatiering.grace.period.millis}} in {{{}DataTieringManager.java{}}}, and modify the {{shouldCacheFile}} logic in {{BucketCache.java}} to respect this grace period for CFs marked accordingly.
Does this approach align with what you had in mind, or am I missing any other part of the flow where the grace period should be applied.
> Allow for a configurable "grace period" when using Time Based Priority
> ----------------------------------------------------------------------
>
> Key: HBASE-29568
> URL: https://issues.apache.org/jira/browse/HBASE-29568
> Project: HBase
> Issue Type: Sub-task
> Reporter: Wellington Chevreuil
> Priority: Major
>
> This aims to introduce a "grace period" functionality in the bucket cache time based priority feature, allowing blocks for the given CF marked for grace period to have its age ignored by the time priority logic.
> The use case here is that "cold" data may have to be accessed rarely, by some batch jobs, so that in these rare job runs, it can be temporarily placed in the cache, then de-prioritised again.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)