[jira] [Commented] (REEF-1142) Wake ProxyObserver caches broken connections

["Andrew Chung (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/REEF-1142?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15120297#comment-15120297 ] 

Andrew Chung commented on REEF-1142:
------------------------------------

[~shulmanb], this seems reasonable. Have you thought about how the retry logic should work? i.e. how many times should the client try to handle the `IOException` before recreating the cached `ILink`, and how many times should the client try to create the new link? It seems that if you would like the user the ability to specify different retry policies for the same endpoint, it may be hard to use the cache at all unless you consider to cache clients based on their retry policies as well. Would a default retry policy be sufficient?

> Wake ProxyObserver caches broken connections
> --------------------------------------------
>
>                 Key: REEF-1142
>                 URL: https://issues.apache.org/jira/browse/REEF-1142
>             Project: REEF
>          Issue Type: Bug
>            Reporter: Boris Shulman
>            Assignee: Boris Shulman
>
> Currently Wake ProxyObserver caches teh actual stream that it gets from Tcp client. Once this stream is broken it impossible to recreate it without recreating RemoteManager



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)