[
https://issues.apache.org/jira/browse/SLIDER-799?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14350635#comment-14350635
]
Steve Loughran commented on SLIDER-799:
---------------------------------------
If we are escalating, we could also consider
# having a rack-local escalation before going cluster wide. This reduces the
cost of fetching blocks from the previous host (assuming it is up & has all the
blocks local).
# maybe even falling back to other labels/queues. This is trickier and could
lead to cluster admins dealing with support problems like 'why is my hbase
master not running on a node of a given label?'. I think I'd rather have the
component request unsatisfied and let those admins add new nodes to the label
set explicitly.
> AM to decide when to relax placement policy from specific host to rack/cluster
> ------------------------------------------------------------------------------
>
> Key: SLIDER-799
> URL: https://issues.apache.org/jira/browse/SLIDER-799
> Project: Slider
> Issue Type: Improvement
> Components: appmaster
> Affects Versions: Slider 0.70
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Priority: Critical
> Fix For: Slider 0.80
>
> Original Estimate: 24h
> Remaining Estimate: 24h
>
> If Slider asks for relaxed affinity, YARN only gives it ~1 second for free
> capacity to appear on a node before it falls back to non-local assignment.
> While this is OK for analytics throughput, it's suboptimal for placement of
> code such as HBase region servers.
> AM needs to take charge of the placement and decide for itself when to
> convert from placed to relaxed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
