Open rrnewton opened 10 years ago
Further increasing the number of threads (e.g. -N10
) makes this happen more often. Also, cranking up STRESSTESTS makes it happen reliably, even with one run of the process.
I repeated the same thing (10K or more runPars) with the old monad-par
package and I didn't have any trouble.
This may have something to do with the other (even simpler) stress test, #81.
Currently, if I run stress tests:
I eventually get this:
Could it be that we're just leaking threads and running into a problem because we have too many? 9000 shouldn't be too many, however.