You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@phoenix.apache.org by "Nick Dimiduk (JIRA)" <ji...@apache.org> on 2016/05/26 01:17:12 UTC
[jira] [Commented] (PHOENIX-2939) MetaCache is easily thrashed with
default settings
[ https://issues.apache.org/jira/browse/PHOENIX-2939?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15301255#comment-15301255 ]
Nick Dimiduk commented on PHOENIX-2939:
---------------------------------------
If we want to continue the practice of caching the entire stats table, we should dynamically resize the metadata cache so that all the stats can fit into the cache.
> MetaCache is easily thrashed with default settings
> --------------------------------------------------
>
> Key: PHOENIX-2939
> URL: https://issues.apache.org/jira/browse/PHOENIX-2939
> Project: Phoenix
> Issue Type: Bug
> Reporter: Nick Dimiduk
>
> With default settings for {{phoenix.coprocessor.maxMetaDataCacheSize}} (20mb) and {{phoenix.stats.guidepost.width}} (100mb * 3), even a relatively small amount of data (ie 10TB, 2000 regions) will easily produce more stats data than fits in the cache. This quickly leads us into a situation where we thrash the cache and spend an inordinate amount of time (re)scanning SYSTEM.STATS table.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)