[jira] [Commented] (AURORA-1871) Client should reject tasks with duplicate process names

2017-03-24 Thread Joshua Cohen (JIRA)

[ 
https://issues.apache.org/jira/browse/AURORA-1871?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15940592#comment-15940592
 ] 

Joshua Cohen commented on AURORA-1871:
--

I don't think we need to do this verification in Pystachio, nor do we need to 
write a parser for the DSL (which Pystachio already is). Instead, we can just 
verify in the client after parsing the config, but before sending the job to 
the scheduler. E.g. in 
[context.py#get_job_config](https://github.com/apache/aurora/blob/master/src/main/python/apache/aurora/client/cli/context.py#L151-L172)
 we could add some additional validation logic.

One thing to keep in mind however, is we'd need to be careful that commands on 
existing jobs that may have invalid config are still possible. We should only 
reject jobs on admission.

> Client should reject tasks with duplicate process names
> ---
>
> Key: AURORA-1871
> URL: https://issues.apache.org/jira/browse/AURORA-1871
> Project: Aurora
>  Issue Type: Task
>  Components: Client
>Reporter: Joshua Cohen
>
> If a user creates a job that contains tasks with the same process name, that 
> info is happily passed on to thermos, which will happily run one of those 
> processes, but maybe display a separate one in the UI. In general the 
> behavior in this case is non-deterministic and can lead to hard to track down 
> bugs.
> We should just short circuit and fail in the client if we detect multiple 
> processes with the same name.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)


[jira] [Commented] (AURORA-1871) Client should reject tasks with duplicate process names

2017-03-24 Thread Pradyumna Kaushik (JIRA)

[ 
https://issues.apache.org/jira/browse/AURORA-1871?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15940560#comment-15940560
 ] 

Pradyumna Kaushik commented on AURORA-1871:
---

1. As pistachio is the one that's parsing the DSL, we would need to change 
pistachio (which doesn't seem the right thing to do) in order to detect 
multiple tasks having the same process name.
2. The other solution to this would be to write a parser for the DSL whose sole 
purpose is to detect flaws in the configuration.

[~jcohen] what do you think about this?

> Client should reject tasks with duplicate process names
> ---
>
> Key: AURORA-1871
> URL: https://issues.apache.org/jira/browse/AURORA-1871
> Project: Aurora
>  Issue Type: Task
>  Components: Client
>Reporter: Joshua Cohen
>
> If a user creates a job that contains tasks with the same process name, that 
> info is happily passed on to thermos, which will happily run one of those 
> processes, but maybe display a separate one in the UI. In general the 
> behavior in this case is non-deterministic and can lead to hard to track down 
> bugs.
> We should just short circuit and fail in the client if we detect multiple 
> processes with the same name.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)


[jira] [Commented] (AURORA-1871) Client should reject tasks with duplicate process names

2017-02-25 Thread Pradyumna Kaushik (JIRA)

[ 
https://issues.apache.org/jira/browse/AURORA-1871?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15884454#comment-15884454
 ] 

Pradyumna Kaushik commented on AURORA-1871:
---

[~joshua.cohen] Just a clarification -- Should the client just fail, in the 
sense that we throw an error indicating a faulty config file or should the job 
with tasks that contain duplicate task names be ignored and the others 
scheduled?

> Client should reject tasks with duplicate process names
> ---
>
> Key: AURORA-1871
> URL: https://issues.apache.org/jira/browse/AURORA-1871
> Project: Aurora
>  Issue Type: Task
>  Components: Client
>Reporter: Joshua Cohen
>
> If a user creates a job that contains tasks with the same process name, that 
> info is happily passed on to thermos, which will happily run one of those 
> processes, but maybe display a separate one in the UI. In general the 
> behavior in this case is non-deterministic and can lead to hard to track down 
> bugs.
> We should just short circuit and fail in the client if we detect multiple 
> processes with the same name.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)