Per Accessor Deadlock Detection

Poster Content
nk4um User
Posts: 3
October 18, 2007 17:37Thanks
Tx Tony.  Looking forward to NK4.
Like · Post Reply
nk4um Administrator
Posts: 607
October 17, 2007 19:28Clarification
sorry that paragraph should have read:

Our experience with deadlock detection has lead us to reconsider it''s inclusion into NetKernel 4 - at least in the kernel. It usually leads to more problems and complexity than the problem it was there to solve. I.e. a guaranteed way on a production server that a coding error wouldn''t lead to complete lock up. NetKernel 4 is architected to allow auxiliary management services to be plugged in around the kernel which would support more flexibility without needing to be built in.
Like · Post Reply
nk4um Administrator
Posts: 607
October 17, 2007 19:26short answer: no
Hi David,

there is no way at the moment to set custom deadlock the periods for different accessors. Your solution is the one that I would have suggested. :-)

Our experience with deadlock detection has lead us to reconsider it''s inclusion into NetKernel 4 - at least in the kernel. It usually leads to more problems and complexity than the problem it was there to solve. I.e. a guaranteed way on a production server that a coding error wouldn''t lead to complete lock up. NetKernel is architected to allow auxiliary management services to be plugged in around the kernel which would support more flexibility without needing to be built in.

So in summary, keep doing what you are doing for the moment.

Cheers,
Tony
Like · Post Reply
nk4um User
Posts: 3
October 17, 2007 16:20Per Accessor Deadlock Detection
Hi Peter & everyone,

Is there any way to set the deadlock detection period on a per accessor basis?  Some of our components are long-running procedures which don''t make sub requests.

My solution to this was to fire off a thread to issue "sleep" requests to NK occasionally for some period longer than the deadlock setting, tailored to the accessor.  This works just fine.  But some of the other guys want an "easier" solution.  :-)

Thanks,
David
Like · Post Reply