hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sunil G (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-2729) Support script based NodeLabelsProvider Interface in Distributed Node Label Configuration Setup
Date Sat, 17 Oct 2015 03:56:05 GMT

    [ https://issues.apache.org/jira/browse/YARN-2729?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14961718#comment-14961718
] 

Sunil G commented on YARN-2729:
-------------------------------

Hi [~Naganarasimha] and [~rohithsharma]
Sorry for coming in late. I have some doubts here.

In {{ScriptBasedNodeLabelsProvider#serviceStop}}, {{shexec}} process is destroyed and the
{{super.stop()}} is invoked. Hence I feel it may still be possible that {{shexec}} may again
get executed (if one timer run happens in this timeframe) before timer is cancelled from {{AbstractNodeLabelsProvider#serviceStop}}.
We can do {{super.stop()}} first from ScriptBasedNodeLabelsProvider. But I feel it may not
be better, so could we reset *shexec* to null after attached process is destroyed (may also
need a null check in timetask if we do this). Thoughts?

couple of other comments:
1. ScriptBasedNodeLabelsProvider#serviceInit: {{verifyConfiguredScript}} is invoked with {{nodeLabelsScript}}
which is a class variable. It may not  be needed.
2. I have one doubt in {{fetchLabelsFromScriptOutput}}. {{scriptOutput}} is split  with newline
and each line is checked whether its starts with NODE_LABEL_PARTITION_PATTERN. But if we have
multiple lines from input string which starts with NODE_LABEL_PARTITION_PATTERN, then only
last line will be set to {{nodePartitionLabel}}. I didnt understand this part. Could you please
have a look on this.

> Support script based NodeLabelsProvider Interface in Distributed Node Label Configuration
Setup
> -----------------------------------------------------------------------------------------------
>
>                 Key: YARN-2729
>                 URL: https://issues.apache.org/jira/browse/YARN-2729
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: nodemanager
>            Reporter: Naganarasimha G R
>            Assignee: Naganarasimha G R
>         Attachments: YARN-2729.20141023-1.patch, YARN-2729.20141024-1.patch, YARN-2729.20141031-1.patch,
YARN-2729.20141120-1.patch, YARN-2729.20141210-1.patch, YARN-2729.20150309-1.patch, YARN-2729.20150322-1.patch,
YARN-2729.20150401-1.patch, YARN-2729.20150402-1.patch, YARN-2729.20150404-1.patch, YARN-2729.20150517-1.patch,
YARN-2729.20150830-1.patch, YARN-2729.20150925-1.patch, YARN-2729.20151015-1.patch, YARN-2729.20151310-1.patch,
YARN-2729.20151310-2.patch
>
>
> Support script based NodeLabelsProvider Interface in Distributed Node Label Configuration
Setup . 
> Miscellaneous Issues :
> # In configurationNodeLabelsProvider instead of taking multiple labels from single configuration,
we need to support exclusive configuration for partition (single label).
> # Proper logging when registration of Node Fails
> # Classloader was not getting reset from custom class loader in TestConfigurationNodeLabelsProvider.java
which could make test cases fail in certain conditions
> # In ResourceTrackerService we need to consider distributed configuration only when node
labels are enabled. leads to lots of logs in certain conditions
> # NodeLabelsProvider needs to be a interface rather than  abstract class 



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

Mime
View raw message