[
https://issues.apache.org/jira/browse/HADOOP-17362?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jim Brennan updated HADOOP-17362:
---------------------------------
Comment: was deleted
(was: Measurements taken by [~daryn] when he made this change internally:
{quote}
Given a real world har for 1 hour of tez jobs from one of our research clusters:
- listing formerly took 12s with ~47.5k rpcs, reduced to 6s with 10 rpcs
- globbing killed after 11min with hundreds of thousands of rpcs with 16G heap,
reduced to 11s with 10 rpcs and 768M heap
{quote})
> Doing hadoop ls on Har file triggers too many RPC calls
> -------------------------------------------------------
>
> Key: HADOOP-17362
> URL: https://issues.apache.org/jira/browse/HADOOP-17362
> Project: Hadoop Common
> Issue Type: Bug
> Components: fs
> Reporter: Ahmed Hussein
> Assignee: Ahmed Hussein
> Priority: Major
> Labels: pull-request-available
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> [~daryn] has noticed that Invoking hadoop ls on HAR is taking too much of
> time.
> The har system has multiple deficiencies that significantly impacted
> performance:
> # Parsing the master index references ranges within the archive index. Each
> range required re-opening the hdfs input stream and seeking to the same
> location where it previously stopped.
> # Listing a har stats the archive index for every "directory". The per-call
> cache used a unique key for each stat, rendering the cache useless and
> significantly increasing memory pressure.
> # Determining the children of a directory scans the entire archive contents
> and filters out children. The cached metadata already stores the exact child
> list.
> # Globbing a har's contents resulted in unnecessary stats for every leaf path.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
