hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Zhijie Shen (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-3051) [Storage abstraction] Create backing storage read interface for ATS readers
Date Wed, 15 Apr 2015 00:04:59 GMT

    [ https://issues.apache.org/jira/browse/YARN-3051?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14495177#comment-14495177

Zhijie Shen commented on YARN-3051:

bq.  Is there a strong reason to alter the uniqueness semantics?

I'm not proposing change the semantics, but according to the db schema design (PK), I'm suspecting
we're changing the semantics implicitly, such that I want to make it clear. In fact, unless
some strong requirement to support the additional use case, I'm incline to keeping the semantics
stable to migrate the existing timeline service users smoothly. Otherwise, it will be difficult
to make {{getEntity(entity type, entity Id)}} compatible.

bq. One could argue that even the addition of cluster id isn't really a change from v.1 as
v.1 didn't envision multi-cluster storage, right?

Yeah, that's another part where I'm a bit confused why we allow the same identifier across
clusters but not across apps. One step back, assume I develop the my framework YTS v2 integration,
due to the constraint of unique entity identifier across apps, I need to carefully define
entity identifier. By doing this, in most cases, the entity identifier is unique across cluster,
thought the assumption of single cluster may break something in the scenario of multiple clusters.
For example, in cluster_1, we will have application_1 and in cluster_2, we will have application_1
too. Having cluster_id can uniquely identify which application_1 it is. I'm just wondering
using cluster_1 to distinguish more than one entities is a rare case if we force users to
define unique entity identifiers across apps, and it sounds more like the cross cluster problem
of this framework. Thoughts?

bq.  We can clearly state that within the same cluster (entity type + entity id) must be unique,
and enforce it within the storage implementation.

+1, we should enforce this constraint. In addition, I think we need to have a index table
such as <entity type, entity id, pointer to the entity in entity table> to support the
single entity query.

> [Storage abstraction] Create backing storage read interface for ATS readers
> ---------------------------------------------------------------------------
>                 Key: YARN-3051
>                 URL: https://issues.apache.org/jira/browse/YARN-3051
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: timelineserver
>            Reporter: Sangjin Lee
>            Assignee: Varun Saxena
>         Attachments: YARN-3051_temp.patch
> Per design in YARN-2928, create backing storage read interface that can be implemented
by multiple backing storage implementations.

This message was sent by Atlassian JIRA

View raw message