[
https://issues.apache.org/jira/browse/MESOS-5545?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15316609#comment-15316609
]
Vinod Kone commented on MESOS-5545:
-----------------------------------
I took a look at the design doc. The primary ideas seem to be 1) implement an
actor inside agent to detect the network topology, 2) allow dynamic update of
topology information and 3) Add `rack_id` to SlaveInfo to influence allocation.
Should 1) be done by the network isolator similar to how the gpu isolator
discovers the network topology? cc [~jieyu] [[email protected]] [~klueska]
I'm not convinced it is straightforward to update the topology online. What
happens to tasks that are already launched with the previous network
configuration in mind? This is no different from the online updating of agent
attributes which is disallowed precisely because of that reason.
3) I'm not convinced that we cannot leverage attributes to communicate topology
information to master and the schedulers; why add first class fields like
`rack_id` to SlaveInfo for this? IIUC the reason for the first class field is
to influence allocator decision? Although I didn't quite understand that part
in the doc. How would this allocation strategy respect DRF?
> Add rack awareness support for Mesos resources
> ----------------------------------------------
>
> Key: MESOS-5545
> URL: https://issues.apache.org/jira/browse/MESOS-5545
> Project: Mesos
> Issue Type: Story
> Components: hadoop, master
> Reporter: Fan Du
>
> Resources managed by Mesos master have no topology information of the
> cluster, for example, rack topology. While lots of data center applications
> have rack awareness feature to provide data locality, fault tolerance and
> intelligent task placement. This ticket tries to investigate how to add rack
> awareness for Mesos resources topology.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
