World Community Grid - View Thread - FreeBSD and lost in running-but-idle results: SMP problem?

World Community Grid Forums

Category: Beta Testing

Forum: Beta Test Support Forum

Thread: FreeBSD and lost in running-but-idle results: SMP problem?

Quick Go »

No member browsing this thread

Thread Status: Active
Total posts in this thread: 3

[ ]

Author

This topic has been viewed 1148 times and has 2 replies

Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline


FreeBSD and lost in running-but-idle results: SMP problem?

I think I've found the correlation.

All machines that can not see results are actually completed are SMP.

I'll reboot one test server which exhibits this problem under uniprocessor kernel for a dozen of hours and followup with the results.

[Sep 23, 2007 6:45:28 AM]

Sekerob
Ace Cruncher
Joined: Jul 24, 2005
Post Count: 20043
Status: Offline


Re: FreeBSD and lost in running-but-idle results: SMP problem?

Me Ausländer do not fully comprehend what you are saying. The WCG sciences on BOINC don't run in SMP, single process, multi threaded. If the systems are configured to do SMP processing only, maybe they could be convinced to do a single threaded science, one at the time, provided WCG recognises the OS you run on as one of the 4 or 5 that are send projects.

BOINC at WCG is able to run multiple sciences on multi core CPU's simultaneous, each bound to a single thread/core. There are me thinking somewhere out there 1 or 2 projects that do SMP, but you'd have to go out and discover. A website collecting info on many projects is this: http://distributedcomputing.info/projects.html

----------------------------------------

WCG

Global & Research > Make Proposal Help: Start Here!
Please help to make the Forums an enjoyable experience for All!

[Sep 23, 2007 9:28:03 AM]

Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline


Re: FreeBSD and lost in running-but-idle results: SMP problem?

There is only one master boinc process, which detects multiple CPUs and forks out multiple subtasks (for now, I have only tho-headed machines, so there are two simultaneously running tasks on each that machine)

On most, if not every, cases tasks are stuck in "Running 100%" state, doen not consume CPU power, and just sleep; suspending the whole project and resuming it after 20-60 seconds usually recovers these tasks to "completed" state, but not always.

[Sep 23, 2007 3:19:50 PM]

[ ]