[
https://issues.apache.org/jira/browse/TEZ-2311?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14639413#comment-14639413
]
Hitesh Shah commented on TEZ-2311:
----------------------------------
Comments:
- The DAGImpl change does not seem right. When vertex A is killed, why is
Vertex B being killed by the DAG? The DAG should be triggering a kill for all
vertices or a sub-set of them on certain conditions. Adding this code creates a
loop of events.
Consider the case where there are 5 vertices: A to E. A fails for some reason.
DAG will trigger kill on B to E. When B returns to DAG as Vertex completed with
state KILLED, DAG will again re-trigger KILL on all vertices. This seems wrong.
VertexImpl changes:
{code}
for (Task task : vertex.tasks.values()) {
{code}
- the new code additions need to do "(vertex.tasks != null &&
vertex.numTasks != 0) " checks to ensure that NEW to KILLED does not cause an
NPE.
> AM can hang if kill received while recovering from previous attempt
> -------------------------------------------------------------------
>
> Key: TEZ-2311
> URL: https://issues.apache.org/jira/browse/TEZ-2311
> Project: Apache Tez
> Issue Type: Bug
> Affects Versions: 0.6.0
> Reporter: Jason Lowe
> Assignee: Jeff Zhang
> Labels: Recovery
> Attachments: TEZ-2311-1.patch
>
>
> We saw an instance of a Tez job hanging despite receiving multiple kill
> requests from clients. The AM was recovering from a prior attempt when the
> first kill request arrived.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
