Craig Welch commented on YARN-1994:

[~mipoto], it's unfortunate that there is that corner case behavior in Windows, 
that doesn't sound correct.  In any case, I can see how you would want this 
behavior (binding to but still specifying a particular address) and 
could think of it as an extension of the single-bind address behavior (in which 
case the address is communicated around just by virtue of the listener binding 
only to it and therefore returning it when requested).  I also don't want to 
have us commit something which doesn't meet your full set of needs.  I was 
concerned that in some cases the original patch changed behavior when bind-host 
was not present, I think our final patch needs to carry forward changes to 
avoid this.  I also think there is value in looking at folding this into 
ipc.Server/Configuration/YarnConfiguration so that the overall flow is as it 
was before, there isn't the re-calculation of address all over 
(re-determination using the address, default address, default port), (to your 
comment about adding it to RPC.Server).  [~arpitagarwal] [~xgong] I have a .9 
patch which is lhe .7 patch with the additional tests and the change to the 
RPCUtil.getSocketAddr but which I never uploaded - you've actually reviewed the 
code already at different times, it's a combination of the current patch and 
.7, effectively.  I would like to take one more look at doing it the way I have 
outlined above, if that proves simple enough I think it will be preferable over 
the long term and will support Milan's case - if not, we can go with the .9 and 
I'll upload it.  Now that I understand it more, I can test it by specifying a 
hostname which is not the host's actual hostname in my test cluster...

> Expose YARN/MR endpoints on multiple interfaces
> -----------------------------------------------
>                 Key: YARN-1994
>                 URL: https://issues.apache.org/jira/browse/YARN-1994
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: nodemanager, resourcemanager, webapp
>    Affects Versions: 2.4.0
>            Reporter: Arpit Agarwal
>            Assignee: Craig Welch
>         Attachments: YARN-1994.0.patch, YARN-1994.1.patch, 
> YARN-1994.11.patch, YARN-1994.11.patch, YARN-1994.12.patch, 
> YARN-1994.2.patch, YARN-1994.3.patch, YARN-1994.4.patch, YARN-1994.5.patch, 
> YARN-1994.6.patch, YARN-1994.7.patch
> YARN and MapReduce daemons currently do not support specifying a wildcard 
> address for the server endpoints. This prevents the endpoints from being 
> accessible from all interfaces on a multihomed machine.
> Note that if we do specify INADDR_ANY for any of the options, it will break 
> clients as they will attempt to connect to We need a solution that 
> allows specifying a hostname or IP-address for clients while requesting 
> wildcard bind for the servers.
> (List of endpoints is in a comment below)

This message was sent by Atlassian JIRA

Reply via email to