hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Steve Loughran (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HADOOP-13550) Azure threaded deleter logs too much at info
Date Fri, 26 Aug 2016 18:35:21 GMT

     [ https://issues.apache.org/jira/browse/HADOOP-13550?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Steve Loughran updated HADOOP-13550:
------------------------------------
    Attachment: HADOOP-13550-branch-2-001.patch

Patch 001; 

* minor cleanup of code, 
* javadocs, 
* cranked back logging
* made all fields final, where possible.
* use hadoop-approved line lengths
* mark threads as daemon

One thing worries me about this code: it creates a thread pool for each operation, but I don't
see that thread pool being shared across operations.

This means that if I set {{"fs.azure.delete.threads"}} to 10, and set of 10 deletes in parallel
on the same JVM, 100 deleter threads will be spun up. That's a risk of running out of resources,
fast. 

Having a pool of threads per FS instance for shared operations would (a) reduce the risk of
resource consumption and (b) eliminate/reduce the time to spin up new threads when an operation
comes in.

> Azure threaded deleter logs too much at info
> --------------------------------------------
>
>                 Key: HADOOP-13550
>                 URL: https://issues.apache.org/jira/browse/HADOOP-13550
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: fs/azure
>    Affects Versions: 2.9.0
>            Reporter: Steve Loughran
>            Priority: Trivial
>         Attachments: HADOOP-13550-branch-2-001.patch
>
>
> The new azure parallel deleter shares too much information in the logs about what is
going on
> {code}
> 2016-08-26 18:56:16,461 WARN  azure.AzureFileSystemThreadPoolExecutor (AzureFileSystemThreadPoolExecutor.java:executeParallel(150))
- Disabling threads for Delete operation as thread count 0 is <= 1
> 2016-08-26 18:56:21,424 INFO  azure.AzureFileSystemThreadPoolExecutor (AzureFileSystemThreadPoolExecutor.java:executeParallel(235))
- Time taken for Delete operation is: 4963 ms with threads: 0
> {code}
> # given that thread count of 0 is the default, warn is overkill and misleading
> # I (and presumably many other users) don't care about how long something took.
> These should both be downgraded to debug



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org


Mime
View raw message