hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Daniel Templeton (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-7237) Cleanup usages of ResourceProfiles
Date Thu, 05 Oct 2017 16:28:00 GMT

    [ https://issues.apache.org/jira/browse/YARN-7237?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16193135#comment-16193135
] 

Daniel Templeton commented on YARN-7237:
----------------------------------------

The changes made look good.  On the open topics:

bq. {quote}Rather than creating our own minimum and maximum profiles based on the usual minimums
and maximums, I rather get rid of the minimum and maximum profiles completely and go back
to the original code for enforcing minimums and maximums.{quote} I would prefer to keep it
so that we can get a completed list of profiles including minimum and maximum.

What do we get from having a completed list of profiles?  Before resource profiles were added,
we dealt with the configured min and max directly.  I don't see where making it a profile
makes the code any cleaner or easier.  In fact, adding another layer of indirection here adds
complication without any clear benefit.  As I'm sure you've noticed by now, I'm very interested
in keeping things as simple as possible and only adding complication when it adds value. 
(Or in this case, removing complication if it doesn't add value.)

If you insist on keeping the minimum and maximum profiles, then you need tests to validate
that the profile names are actually disallowed.

bq. {quote}testResourceProfiles() looks like it tests more than just the minimum and maximum
profiles. Are you sure you should remove it?{quote}I'm not quite understand this comment,
could you explain?

In the patch you removed {{testResourceProfiles()}}, but it looks like the test was testing
more than just the minimum and maximum profiles.  My question is whether it might not be better
to just remove the minimum and maximum profile portions of the test rather than removing the
whole test.

> Cleanup usages of ResourceProfiles
> ----------------------------------
>
>                 Key: YARN-7237
>                 URL: https://issues.apache.org/jira/browse/YARN-7237
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: nodemanager, resourcemanager
>            Reporter: Wangda Tan
>            Assignee: Wangda Tan
>            Priority: Critical
>         Attachments: YARN-7237.001.patch, YARN-7237.002.patch, YARN-7237.003.patch, YARN-7237.004.patch,
YARN-7237.005.patch
>
>
> While doing tests, there're a couple of issues:
> 1) When use {{ProfileCapability#getProfileCapabilityOverride}}, it does overwrite of
whatever specified in resource-profiles.json when value >= 0. Which is different from javadocs
of {{ProfileCapability}} 
> bq. For example, if you have a resource profile "small" that maps to <4096M, 2 cores,
1 gpu> and you set the capability override to <8192M, 0 cores, 0 gpu>, then the actual
resource allocation on the ResourceManager will be <8192M, 2 cores, 1 gpu>
> To me, the correct behavior should do overwrite when value > 0. The reason is, by
default resource value will be set to 0, For example, assume we have a profile {{"a" = (mem=3,
vcore=5, res_1=7)}}, and create a capability-overwrite (capability = new resource(8). The
final result should be (mem=8, vcore=5, res_1=7), instead of (mem=8, vcore=0, res_1=0).
> 2) ResourceProfileManager now loads minimum/maximum profile from config file (resource-profiles.json),
to me this is not correct because minimum/maximum allocation for each resource types are already
specified inside {{resource-types.xml}}. We should always use {{ResourceUtils#getResourceTypesMinimum/MaximumAllocation}}
to get from resource-types.xml and yarn-site.xml. This value will be added to profiles so
client can get these configs.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

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


Mime
View raw message