The server tells the client to abort a result only when 1) the result isn't in the DB 2) the result's WU was cancelled. In either case, it writes a string of the form "[HOST#%d]: Send result_abort for result %s; reason: %s\n" to the scheduler log.
So scan your scheduler log and see if you're getting any of these. -- David Rytis Slatkevičius wrote: > It seems that BOINC sometimes issues server-side task kill when it isn't > really needed, aborting WUs in progress. I have one example here: > http://www.primegrid.com/workunit.php?wuid=95082238 > > <http://www.primegrid.com/workunit.php?wuid=95082238>The task was humming > along nicely, when it got killed from server side (marked as redundant > result) and aborted on client, wasting CPU time. Seconds later, a new task > was created and sent to another client. This isn't an isolated issue either, > I have several more WUs that I know of, all resulting in a significant waste > of CPU time. > > The server is running revision 19870. > _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
