[
https://issues.apache.org/jira/browse/EAGLE-310?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Michael Wu updated EAGLE-310:
-----------------------------
Description:
Preproduce:
1. create a topology and an execution bound to it, and then start the execution.
2. deleted the execution started in the last step without stopping it.
3. recreate the topology and execution, with same topology name.
4. as a result, on the monitoring page, the recreated execution displays "NEW"
status that customers can hit the "start" button and get an error message
"Topology with name `eagle-sandbox-hbaseSecurityLog-demo001` already exists on
cluster", however, the execution status is refreshed to "STARTED" in 5 seconds
after syncing with the cluster.
Investigated cause: as deleting topology and execution without stopping the
execution, system doesn't actually delete the corresponding topology in storm,
so when we recreate the pair (topology+execution), the same named topology
cannot be recreated in storm.
Point of view: customers will be confused if they recreate the same named pair,
with error message but get it running in 5 seconds, meanwhile, being able to
hit "start" button of execution will make them confused also.
Action required: please consider, as a user recreate a same named execution, if
it's possible to get topology status in storm before showing the list on
monitoring page, and show the real status of the execution in the list after
that, as well as changing the "start" button to "stop" button accordingly.
There should be a notification issue of deleting a "started" execution, please
refer: https://issues.apache.org/jira/browse/EAGLE-312
was:
Preproduce:
1. create a topology and an execution bound to it, and then start the execution.
2. deleted the execution started in the last step without stopping it.
3. recreate the topology and execution, with same topology name.
4. as a result, on the monitoring page, the recreated execution displays "NEW"
status that customers can hit the "start" button and get an error message
"Topology with name `eagle-sandbox-hbaseSecurityLog-demo001` already exists on
cluster", however, the execution status is refreshed to "STARTED" in 5 seconds
after syncing with the cluster.
Investigated cause: as deleting topology and execution without stopping the
execution, system doesn't actually delete the corresponding topology in storm,
so when we recreate the pair (topology+execution), the same named topology
cannot be recreated in storm.
Point of view: customers will be confused if they recreate the same named pair,
with error message but get it running in 5 seconds, meanwhile, being able to
hit "start" button of execution will make them confused also.
Action required: please consider, as a user recreate a same named execution, if
it's possible to get topology status in storm before showing the list on
monitoring page, and show the real status of the execution in the list after
that, as well as changing the "start" button to "stop" button accordingly.
There should be a notification issue of deleting a "started" execution, please
refer:
> already existing active topology status not displayed when a deleted
> topology+execution re-created with same name
> -----------------------------------------------------------------------------------------------------------------
>
> Key: EAGLE-310
> URL: https://issues.apache.org/jira/browse/EAGLE-310
> Project: Eagle
> Issue Type: Bug
> Affects Versions: v0.3.0
> Reporter: Michael Wu
> Assignee: Zhao, Qingwen
> Fix For: v0.4.0
>
>
> Preproduce:
> 1. create a topology and an execution bound to it, and then start the
> execution.
> 2. deleted the execution started in the last step without stopping it.
> 3. recreate the topology and execution, with same topology name.
> 4. as a result, on the monitoring page, the recreated execution displays
> "NEW" status that customers can hit the "start" button and get an error
> message "Topology with name `eagle-sandbox-hbaseSecurityLog-demo001` already
> exists on cluster", however, the execution status is refreshed to "STARTED"
> in 5 seconds after syncing with the cluster.
> Investigated cause: as deleting topology and execution without stopping the
> execution, system doesn't actually delete the corresponding topology in
> storm, so when we recreate the pair (topology+execution), the same named
> topology cannot be recreated in storm.
> Point of view: customers will be confused if they recreate the same named
> pair, with error message but get it running in 5 seconds, meanwhile, being
> able to hit "start" button of execution will make them confused also.
> Action required: please consider, as a user recreate a same named execution,
> if it's possible to get topology status in storm before showing the list on
> monitoring page, and show the real status of the execution in the list after
> that, as well as changing the "start" button to "stop" button accordingly.
> There should be a notification issue of deleting a "started" execution,
> please refer: https://issues.apache.org/jira/browse/EAGLE-312
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
