On Thu, 27 May 2010 16:09:59 -0400, you wrote:

>One of our users has appeared to have found a way to get around the workunit 
>cache limit.  I was wondering if there's anything we can do about this.  
>
>milky...@home has an 8 WU per core limit, and for some reason this user:
>
>http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=171705
>
>has around 3000 tasks pending, which shouldn't be possible.
>
>Is this because of some bug, or malicious use?  It really screws up some of 
>the validation as a lot of those workunits are part of a quorum.  
>

Quick test, grab some work, reset project, connect to Milkyway again.

Project haven't turned-on re-issue of "lost" wu, so client gets new work
instead.

So basically, user doesn't need to do anything wrong, but for some of very many
reasons, the scheduler-reply doesn't make it successfully back to client.

For the user, if he looks on BOINC-client, the most likely result is he'll see
client asked for work, but the connection timed-out.

For project, turning-on re-issue of "lost" work will stop this problem from
happening. But of course, it's possible due to database-performance that this
will add too much extra load so isn't an option...

Oh, and in case someone thinks I've now got a bunch of work that will time-out
later-on, no problem, a detach & re-attach does work, and all work is marked
"client detached". ;)




-- 
"I make so many mistakes. But then just think of all the mistakes
I don't make, although I might."
_______________________________________________
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.

Reply via email to