[
https://issues.apache.org/jira/browse/HADOOP-13649?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16446485#comment-16446485
]
Aaron Fabbri commented on HADOOP-13649:
---------------------------------------
Assigning to [~gabor.bota]. Feel free to change the datastructures used. The
"two hashtables" was a quick and dirty design that may not be the best if we
add TTL-based expiry. I would look into [~ste...@apache.org]'s suggestion to
use Guava cache as well. Not sure if it has the needed functionality to
support prune and all the other operations that go with it (note that prune()
sometimes needs to clear isAuthorititative bit on a directory if a file in that
directory is pruned).
> s3guard: implement time-based (TTL) expiry for LocalMetadataStore
> -----------------------------------------------------------------
>
> Key: HADOOP-13649
> URL: https://issues.apache.org/jira/browse/HADOOP-13649
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Affects Versions: 3.0.0-beta1
> Reporter: Aaron Fabbri
> Assignee: Aaron Fabbri
> Priority: Minor
>
> LocalMetadataStore is primarily a reference implementation for testing. It
> may be useful in narrow circumstances where the workload can tolerate
> short-term lack of inter-node consistency: Being in-memory, one JVM/node's
> LocalMetadataStore will not see another node's changes to the underlying
> filesystem.
> To put a bound on the time during which this inconsistency may occur, we
> should implement time-based (a.k.a. Time To Live / TTL) expiration for
> LocalMetadataStore
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
