leaking blocked processes in vmstat ... how to debug?

[Chuck Swiger]

Marc G. Fournier wrote:
> Up to 48 right now, and still nothing to show for it ...

Perhaps it is not clear to you that under Unix is it normal for processes to 
block waiting on events like user input, a signal indicating a timer has fired 
or a pipe has more data, etc.

The reason you are looking for processes stuck in "D" for "Disk wait" state is 
to identify any large processes that are being swapped out and thus being held 
up by RAM limitations.  On the other hand, it's entirely possible that you've 
simply got too much stuff running for the available hardware, and will need to 
either add RAM or offload some of the workload to other machines.

-- 
-Chuck