streams-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Renato Marroquín Mogrovejo <renatoj.marroq...@gmail.com>
Subject Re: Exposing Bugs
Date Sat, 20 Sep 2014 12:26:48 GMT
Sorry Ryan I sent that too fast I saw you created STREAMS-179.
Thanks! I will try debugging RSS provider with that :)

Renato M.
On Sep 20, 2014 3:16 AM, "Renato Marroquín Mogrovejo" <
renatoj.marroquin@gmail.com> wrote:

> These are great news Ryan! So will you update us when the pull request is
> up? Are you going to create a JIRA for keeping track of this?
> I am really eager to get the RSS provider working :-)
>
> Renato M.
> On Sep 19, 2014 10:44 PM, "Ryan Ebanks" <ryanebanks@gmail.com> wrote:
>
>> I have issued a pull request that allows local streams to  shutdown if a
>> provider, processor, or writer tasks stops execution due to an unhandled
>> throwable.  This should be a step forward in making streams more
>> transparent and reliable on the local level.  The Throwable will appear in
>> the logs and it will note which Runnable class threw the error.
>>
>> This is intended to only be a temporary fix.  This will illuminate many
>> streams operations that have bugs, and need better testing.  Please create
>> tickets for all exceptions that are caused by bugs and issue pull requests
>> (with appropriate tests and documentation) to fix them.
>>
>> In the future we can relaunch threads that exit due to error, and
>> reprocess
>> the failed datum.  Until we implemented better datum tacking this is not
>> possible, but it is on the road map.
>>
>> Please create tickets or notify the dev list of bugs you encounter.  This
>> push should expose bugs that previously were causing streams to fail
>> silently or stall/hang.  We need to have big push for better tested code
>> in
>> the project.
>>
>> Note: Threads that exit due to error that are not started/handled by the
>> local stream builder will still fail silently.  The fact that so many
>> providers, processors, and writers have to launch their own threads means
>> we should work on the local frame work to remove those needs push almost
>> all thread management responsibility to the framework so we can ensure
>> consistent monitoring/error handling of threads.
>>
>> Thanks,
>> Ryan Ebanks
>>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message