| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: No new tasks in quite a while |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 46
|
|
| Author |
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7219 Status: Offline Project Badges: 
|
At this time my Windows system has a full queue of 20. My Linux systems have a queue a big fat zero. MCM only. No GPU's
----------------------------------------Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Grumpy Swede
Master Cruncher Svíþjóð Joined: Apr 10, 2020 Post Count: 1866 Status: Offline Project Badges: 
|
I get the few MCM task my 8 threaded Windows computer need (I only run 4 threads, and only the WCG project) On top of that I ask for 6 more MCM tasks to keep in the cache. So far, it has never gone completely dry. OPNG on the other hand, seems as rare as hen's teeth now. I haven't seen more than 4 OPNG tasks the last 24 hours.
----------------------------------------  |
||
|
|
AlmightyMoe
Cruncher Joined: Nov 2, 2016 Post Count: 31 Status: Offline Project Badges: 
|
Never knew WU were assigned by platform or platform dependent. Interesting. Disheartening as I only use linux for efficiency reasons, but interesting at least. See this thread.Cheers, Mark Thanks. This talks about verifying WUs from the same OS. It doesn't mention anything as far as I read that WUs are assigned to specific OSs to start. Is there a thread that goes deeper on this?  |
||
|
|
adriverhoef
Master Cruncher The Netherlands Joined: Apr 3, 2009 Post Count: 1978 Status: Offline Project Badges: 
|
AlmightyMoe asked if tasks/WUs are assigned to specific OSs to start and if there is a thread that goes deeper on this.
knreed, former World Community Grid Tech, wrote something about this in post 450495: On the server, there is a daemon called the 'feeder' which loads more tasks into a shared memory segment every second. The shared memory segment can store up to 1000 tasks. When the scheduler receives a request from a client to fetch new tasks to work on, the scheduler only looks at this shared memory segment for tasks that it considers for sending. This is done because at any given time there are up to 1 million tasks available to send for a project. If the scheduler hit the database directly to search for tasks to send on each request, then the load on the database would be overwhelming. Note the difference between tasks and WUs (workunits). A workunit can consist of several tasks. A client will receive tasks, not workunits. A task can consist of one or more jobs, especially in the case of the GPU-app 'OPNG'.The shared memory is divided into 'slots' that are allocated to the different projects that we have active. Only a task for that project can be assigned to a specific slot. knreed continues: There are certain times each day when the slots for a given project in the shared memory segment becomes "full" of tasks that are designated for a particular platform (Linux or Windows). These usually only last a few minutes or seconds, but can sometimes last up to 30 minutes. If for example, you are running Linux and the slots assigned to MCM1 are full of tasks that are already assigned to Windows, then the server will treat it as no work is available for MCM1 for you. (This is what triggers the 'tasks are assigned to another platform' message). If you have the 'send other' work flag checked, it will then look for other projects to send you work from. Also, SekeRob wrote in post 548494: "There's no such thing as pre-assignment of work blocks to a platform". Other interesting reads are SekeRob's post 526575: "Rule of thumb... keep your Minimum Buffer to -less- than half of the shortest deadline and keep the MaxAB [Max Additional workBuffer] to zero.", post 548523: "Principally, task A or B or C does not get pre-assigned to a platform" and "Key is to NOT use the MaxAB function. Just MinB, for that forces connection to keep the buffer up." Adri |
||
|
|
alanb1951
Veteran Cruncher Joined: Jan 20, 2006 Post Count: 729 Status: Offline Project Badges: 
|
Never knew WU were assigned by platform or platform dependent. Interesting. Disheartening as I only use linux for efficiency reasons, but interesting at least. See this thread.Cheers, Mark Thanks. This talks about verifying WUs from the same OS. It doesn't mention anything as far as I read that WUs are assigned to specific OSs to start. Is there a thread that goes deeper on this? If "Homogeneous redundancy" (HR) is used for an application, the HR class for a specific WU is not set until the first task for that WU has been allocated to a host. It is possible to create jobs intended for a specific HR class, but I don't think WCG runs like that... If you are interested, you might look at Homogeneous Redundancy documentation; relevant pages at the original BOINC Wiki or the newer BOINC WIki at GitHub Cheers - Al. P.S. I see Adri was composing an answer at the same time I was :-) -- interesting bit of in-depth work there!... [Edited to add the postscript!] [Edit 1 times, last edit by alanb1951 at Apr 10, 2024 11:30:19 AM] |
||
|
|
gibbcorp
Advanced Cruncher Joined: Nov 29, 2005 Post Count: 72 Status: Offline Project Badges: 
|
Not sure if this is the right place to post but i will anyway. I have a Mac Mini that has been getting hardly any work for a couple of months now. It used to be doing 70k points per day and now does about 5000. Most of the time it is waiting for tasks. Is this going to change? If not I might sell it. thanks
|
||
|
|
BobbyB
Veteran Cruncher Canada Joined: Apr 25, 2020 Post Count: 575 Status: Offline Project Badges:
|
What WCG projects are you subscribed to in your profile?
These days it seems to be all MCM1 and my queues are filled for the day. But mine are not a Mac mini. Don't know what they can do. My profiles allow all projects. |
||
|
|
gibbcorp
Advanced Cruncher Joined: Nov 29, 2005 Post Count: 72 Status: Offline Project Badges:
|
I'm subscribed to all of them.
|
||
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 1842 Status: Offline Project Badges:
|
Not sure if this is the right place to post but i will anyway. I have a Mac Mini that has been getting hardly any work for a couple of months now. It used to be doing 70k points per day and now does about 5000. Most of the time it is waiting for tasks. Is this going to change? If not I might sell it. thanks Have one older MacBook Pro running with macOS and that has not missed a beat once the basic project issues had been resolved, don't think that it ever even got close to running even low on WUs. Ralf  |
||
|
|
BobbyB
Veteran Cruncher Canada Joined: Apr 25, 2020 Post Count: 575 Status: Offline Project Badges:
|
Between when it was doing 70k points per day and now doing about 5000, has anything changed? What is the work that it does get now. I'm just asking questions to get the thinking going. I have no hunch.
|
||
|
|
|