[
https://issues.apache.org/jira/browse/FLINK-5098?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15676758#comment-15676758
]
ASF GitHub Bot commented on FLINK-5098:
---------------------------------------
GitHub user tillrohrmann opened a pull request:
https://github.com/apache/flink/pull/2830
[FLINK-5098] [akka] Detect unreachable remote actors to fail ask calls
eagerly
This PR adds to every sent ask message an additional Identify message which
is used to
detect whether the target actor is actually reachable. The Identify message
allows to
detect unreachable ActorSystems or that an actor is not existing in an
ActorSystem without
having to wait for a timeout. This then allows to fail the ask operation
eagerly.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/tillrohrmann/flink eagerTimeout
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/2830.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #2830
----
commit 0e4f059ca2a1764082851c8e7538483180c3e40b
Author: Till Rohrmann <[email protected]>
Date: 2016-11-09T14:29:32Z
[FLINK-5098] [akka] Detect unreachable remote actors to fail ask calls
eagerly
This PR adds to every sent ask message an additional Identify message which
is used to
detect whether the target actor is actually reachable. The Identify message
allows to
detect unreachable ActorSystems or that an actor is not existing in an
ActorSystem without
having to wait for a timeout. This then allows to fail the ask operation
eagerly.
----
> Detect network problems to eagerly time out ask operations
> ----------------------------------------------------------
>
> Key: FLINK-5098
> URL: https://issues.apache.org/jira/browse/FLINK-5098
> Project: Flink
> Issue Type: Improvement
> Components: Distributed Coordination
> Affects Versions: 1.2.0
> Reporter: Till Rohrmann
> Assignee: Till Rohrmann
> Fix For: 1.2.0
>
>
> Akka's ask operations are given a timeout after which they should fail with
> an {{AskTimeoutException}}. In some cases, however, it is possible to fail
> early because one has detected that the remote host is not reachable or that
> the actor does not exist on the remote {{ActorSystem}}.
> Usually failing early if one cannot hope for a successful message delivery is
> a desirable behaviour since it speeds up recovery.
> I propose to send Akka's {{Identify}} message with each ask request. The
> identify message allows to detect unreachable/non-existing actors and, thus,
> enables us to fail the ask operation early.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
