[ 
https://issues.apache.org/jira/browse/PIG-2032?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Olga Natkovich updated PIG-2032:
--------------------------------

      Description: 
This is from a mail I sent Ben and Chris in feb
---

Essentially the changes are :

a) Allow for progress reporting so that long running agent api invocations dont 
result in killing the task.

b) Decouple creation from initialization of monitor agent harness.

c) I increased the threads for coordinator since ibis has slightly io bound 
coordinator messages to be processed : but this might not be generally relevant.
Maybe some way to configure it would be good ? (The default value was observed 
to be exhausted really fast for some invocations of ibis !)


d) Handle shutdown more gracefully at the agent side - monitor and wait for 
relevant future's to be complete.
Note that the code for this is slightly sensitive in terms of how it is written 
since there are idioms in netty which it supports (immediate invocation vs 
deferred invocation on future completion).



  was:

This is from a mail I sent Ben and Chris in feb
---

Essentially the changes are :

a) Allow for progress reporting so that long running agent api invocations dont 
result in killing the task.

b) Decouple creation from initialization of monitor agent harness.

c) I increased the threads for coordinator since ibis has slightly io bound 
coordinator messages to be processed : but this might not be generally relevant.
Maybe some way to configure it would be good ? (The default value was observed 
to be exhausted really fast for some invocations of ibis !)


d) Handle shutdown more gracefully at the agent side - monitor and wait for 
relevant future's to be complete.
Note that the code for this is slightly sensitive in terms of how it is written 
since there are idioms in netty which it supports (immediate invocation vs 
deferred invocation on future completion).



    Fix Version/s:     (was: 0.9.0)

Since we already branch for 0.9, I don't believe and penny improvements will 
make it in. 

In general, for improvements to a particular feature to make it to a particular 
release it is good to have resources lined to work on the feature.

> Penny related multiple improvements/issues.
> -------------------------------------------
>
>                 Key: PIG-2032
>                 URL: https://issues.apache.org/jira/browse/PIG-2032
>             Project: Pig
>          Issue Type: Bug
>          Components: tools
>    Affects Versions: 0.9.0
>         Environment: All env
>            Reporter: Mridul Muralidharan
>         Attachments: diffs
>
>
> This is from a mail I sent Ben and Chris in feb
> ---
> Essentially the changes are :
> a) Allow for progress reporting so that long running agent api invocations 
> dont result in killing the task.
> b) Decouple creation from initialization of monitor agent harness.
> c) I increased the threads for coordinator since ibis has slightly io bound 
> coordinator messages to be processed : but this might not be generally 
> relevant.
> Maybe some way to configure it would be good ? (The default value was 
> observed to be exhausted really fast for some invocations of ibis !)
> d) Handle shutdown more gracefully at the agent side - monitor and wait for 
> relevant future's to be complete.
> Note that the code for this is slightly sensitive in terms of how it is 
> written since there are idioms in netty which it supports (immediate 
> invocation vs deferred invocation on future completion).

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to