[Openais] Re: saEvtPublish try again

Mark Haverkamp markh at osdl.org
Tue Feb 15 14:50:58 PST 2005


On Tue, 2005-02-15 at 15:46 -0700, Steven Dake wrote:
> Mark,
> 
> I have been working on defect-188 and looking through the library code. 
> I have one comment and suggest we change how saEvtPublish works.
> 
> Today saEvtPublish will retry a publish command and return
> SA_AIS_ERR_TIMEOUT if it receives SA_AIS_ERR_TRYAGAIN from the executive
> too many times.
> 
> Unfortunately this uses nanosleep which only has 1 msec resolution on
> 2.6, and 10msec resolution 2.4.
> 
> I suggest we remove the retry logic and instead let saEvtPublish just
> return SA_ERR_TRYAGAIN and let the application deal with it.  If the
> application wants to drop the message it can, or it can choose to retry.

I could do this now.  The A-spec didn't specify SA_ERR_TRY_AGAIN as one
of the valid return codes for the publish function.  I just looked and
the B-spec does have it specified.  I'll go ahead and make that change.

I am planning to go through the return codes for each of the functions
in the B-spec and make sure that what is being returned is OK.


Mark.


> 
> This is the model by which flow control is expected to occur within the
> ais spec and what is used for checkpointing and other services.
> 
> Your thoughts?
> -steve
-- 
Mark Haverkamp <markh at osdl.org>




More information about the Openais mailing list