[
https://issues.apache.org/jira/browse/FLUME-2502?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Prateek Rungta updated FLUME-2502:
----------------------------------
Attachment: FLUME-2502-1.patch
Adding new patch which consolidates "BATCH" and "RANDOM" consumption orders.
Re your comments:
- The new code uses the same filter as the existing code base for retrieving
file listing so no issue there.
- If the File does not exist after it has been listed from the directory:
The code base to handle it is unchanged from what was done earlier. My
understanding is the `openFile` method will handle the errors as you indicated
- i.e. log an error and move on.
> Spool source's directory listing is inefficient
> -----------------------------------------------
>
> Key: FLUME-2502
> URL: https://issues.apache.org/jira/browse/FLUME-2502
> Project: Flume
> Issue Type: Improvement
> Components: Sinks+Sources
> Affects Versions: v1.5.0
> Reporter: Prateek Rungta
> Attachments: FLUME-2502-0.patch, FLUME-2502-1.patch
>
>
> As mentioned in
> [FLUME-2309|https://issues.apache.org/jira/browse/FLUME-2309], the directory
> listing can it self become the bottleneck when accessing directories with a
> large number of files (>1M). The fix in that JIRA added in the ability to
> specify `RANDOM` as a Consume-Order to avoid sorting large lists.
> The slowness of the directory listing is still un-addressed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
