[
https://issues.apache.org/jira/browse/AURORA-260?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chris Lambert updated AURORA-260:
---------------------------------
Description:
Completely killing a job is a relatively rare operation, and one that should
not be taken lightly. We have seen instances where users intend to invoke
{{aurora kill}} with the {{--shards}} argument, but for various reasons may
wind up killing the entire job.
It's a prudent for us to impose a barrier for a user to back out of a
destructive action before it's too late, and also to distinguish partial job
kills from complete job kills.
I propose we introduce a new command: {{aurora killall}}, which is the means of
completely killing a job (killing all instances of a job). Additionally,
{{aurora kill}} will require a specification of shards/instances. -I propose
that the new {{killall}} subcommand pause for a yet-to-be-determined period
after displaying a warning that the entire job is about to be killed. This
would allow the user time to notice an unintentionally-broad action. Users
wanting to immediately kill an entire job (without the pause) would be forced
to explicitly do so by specifying all instance IDs to the {{kill}} command.-
was:
Completely killing a job is a relatively rare operation, and one that should
not be taken lightly. We have seen instances where users intend to invoke
{{aurora kill}} with the {{--shards}} argument, but for various reasons may
wind up killing the entire job.
It's a prudent for us to impose a barrier for a user to back out of a
destructive action before it's too late, and also to distinguish partial job
kills from complete job kills.
I propose we introduce a new command: {{aurora killall}}, which is the means of
completely killing a job (killing all instances of a job). Additionally,
{{aurora kill}} will require a specification of shards/instances. I propose
that the new {{killall}} subcommand pause for a yet-to-be-determined period
after displaying a warning that the entire job is about to be killed. This
would allow the user time to notice an unintentionally-broad action. Users
wanting to immediately kill an entire job (without the pause) would be forced
to explicitly do so by specifying all instance IDs to the {{kill}} command.
> Improve safety of killing a job
> -------------------------------
>
> Key: AURORA-260
> URL: https://issues.apache.org/jira/browse/AURORA-260
> Project: Aurora
> Issue Type: Story
> Components: Client
> Reporter: Bill Farner
> Assignee: Mark Chu-Carroll
>
> Completely killing a job is a relatively rare operation, and one that should
> not be taken lightly. We have seen instances where users intend to invoke
> {{aurora kill}} with the {{--shards}} argument, but for various reasons may
> wind up killing the entire job.
> It's a prudent for us to impose a barrier for a user to back out of a
> destructive action before it's too late, and also to distinguish partial job
> kills from complete job kills.
> I propose we introduce a new command: {{aurora killall}}, which is the means
> of completely killing a job (killing all instances of a job). Additionally,
> {{aurora kill}} will require a specification of shards/instances. -I propose
> that the new {{killall}} subcommand pause for a yet-to-be-determined period
> after displaying a warning that the entire job is about to be killed. This
> would allow the user time to notice an unintentionally-broad action. Users
> wanting to immediately kill an entire job (without the pause) would be forced
> to explicitly do so by specifying all instance IDs to the {{kill}} command.-
--
This message was sent by Atlassian JIRA
(v6.2#6252)
