[
https://issues.apache.org/jira/browse/FLINK-14452?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zhu Zhu updated FLINK-14452:
----------------------------
Description:
Currently there are 3 failover topology instances created, 2 permanently kept
in JM. 2 scheduling topology instances created permanently kept in JM. It a
waste of computation to build the topologies and memory to keep these
topologies. Which may be a significant issue when the job scale is large.
With FLINK-14450 and FLINK-14451, the SchedulingTopology and FailoverTopology
are able to share one default implementation.
We can change the scheduler to create and keep only one such an execution
topology instance to reduce the cost to build and host execution topologies.
More details see FLINK-14330 and the [design
doc|https://docs.google.com/document/d/1f88luAOfUQ6Pm4JkxYexLXpfH-crcXJdbubi1pS2Y5A/edit#].
was:
With FLINK-14450 and FLINK-14451, the SchedulingTopology and FailoverTopology
are able to share one default implementation.
Currently there are 3 failover topology instances created, 2 permanently kept
in JM. 2 scheduling topology instances created permanently kept in JM. This can
be a waste of
We can change the scheduler to create and keep only one such an execution
topology instance to reduce the cost to build and host execution topologies.
More details see FLINK-14330 and the [design
doc|https://docs.google.com/document/d/1f88luAOfUQ6Pm4JkxYexLXpfH-crcXJdbubi1pS2Y5A/edit#].
> Keep only one execution topology in scheduler
> ---------------------------------------------
>
> Key: FLINK-14452
> URL: https://issues.apache.org/jira/browse/FLINK-14452
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Coordination
> Affects Versions: 1.10.0
> Reporter: Zhu Zhu
> Assignee: Zhu Zhu
> Priority: Major
> Fix For: 1.10.0
>
>
> Currently there are 3 failover topology instances created, 2 permanently kept
> in JM. 2 scheduling topology instances created permanently kept in JM. It a
> waste of computation to build the topologies and memory to keep these
> topologies. Which may be a significant issue when the job scale is large.
> With FLINK-14450 and FLINK-14451, the SchedulingTopology and FailoverTopology
> are able to share one default implementation.
> We can change the scheduler to create and keep only one such an execution
> topology instance to reduce the cost to build and host execution topologies.
> More details see FLINK-14330 and the [design
> doc|https://docs.google.com/document/d/1f88luAOfUQ6Pm4JkxYexLXpfH-crcXJdbubi1pS2Y5A/edit#].
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
