You need to allow test applications in your preferences and you should probably disable FGRPB1 so that the locality scheduler can assign work more easily. I also noticed a potential problem with the scheduler that I need to investigate a bit more. On my desktop I only got one GW task and now there are three cores idle.
You need to allow test applications in your preferences and you should probably disable FGRPB1 so that the locality scheduler can assign work more easily.
Test applications (or beta) has been enabled for months. And FGRPB is disabled. Still no GW work.
I'll soon have to re-enable FGRPB or run out of CPU work all togheter
You need to allow test applications in your preferences and you should probably disable FGRPB1 so that the locality scheduler can assign work more easily. I also noticed a potential problem with the scheduler that I need to investigate a bit more. On my desktop I only got one GW task and now there are three cores idle.
I noticed an entry in the server log (for Arecibo GPUs, but I think it was generic): "only one Beta task per workunit". If that's in effect, and there are no non-Beta apps yet, then the WUs will never have a validation partner.
You need to allow test applications in your preferences and you should probably disable FGRPB1 so that the locality scheduler can assign work more easily. I also noticed a potential problem with the scheduler that I need to investigate a bit more. On my desktop I only got one GW task and now there are three cores idle.
I noticed an entry in the server log (for Arecibo GPUs, but I think it was generic): "only one Beta task per workunit". If that's in effect, and there are no non-Beta apps yet, then the WUs will never have a validation partner.
That is only the case when there is a non-beta app_version and a beta app_version (as is the case for BRP4G). The Beta flag is set for all GW app_versions so they will validate against each other (once I test and activate the validator).
There seems to be something wrong with the scheduler which I will investigate once I fixed the problem with the datafiles.
I had a hard time getting the beta too. I had to request it three or four times, even though I had four empty cores and nothing else running. Finally, I picked up one work unit, and then several more followed on subsequent requests. I ended up with 18, of which 14 completed without errors and the other 4 errored out (BOINC 7.6.33, Win7 64-bit, i7-4771).
I have had Test Applications selected all the time, I dis-abled Gamma-Ray work to try and get the new work units.
It seems that something has worked, I managed to get 4 Multi-Directed work units a couple of hours ago.
Only on my Windows machine so far but I will wait.
A curious thing that I have noticed is that the Gamma-Ray work units state they are "Beta" in their name, but the Multi-Directed work units do not state they are Beta so is the tuning done and these are production work units?
I just aborted about 300 GW as there was no way for my machines (i7s 6950X, 6900, 5960X all OC to 4.0GHz) to process them all in less than 72 hours. I'm finding that if all CPU cores are engaged that work is taking anywhere from 9-11 hours for the longest work units. Haven't seen any validations yet.
I removed the Beta flag on the applications. Everyone should get some GW work now. I'm still monitoring the scheduler and may start to take a look at the results waiting for validation.
I'll be watching the project over the weekend from time to time in case something else breaks.
The datafile issue is now understood and only affects a small range of tasks which I'm going to cancel once they show symptoms.
Of the seven tasks sent to one of my hosts within a few seconds of each other about 12 Oct 2016, 22:42:30 UTC, it remains true that all seven show the second task as unsent at this time. For example this one.
On another host, a single one of these tasks did get the second task sent to a partner, who returned his a while after me, but still over half a day ago, at 14 Oct 2016, 2:21:52 UTC. This WU is perhaps waiting for a manually started run of the validator, as I suppose that is not running automatically just yet.
You need to allow test
)
You need to allow test applications in your preferences and you should probably disable FGRPB1 so that the locality scheduler can assign work more easily. I also noticed a potential problem with the scheduler that I need to investigate a bit more. On my desktop I only got one GW task and now there are three cores idle.
Christian Beer wrote:You need
)
Test applications (or beta) has been enabled for months. And FGRPB is disabled. Still no GW work.
I'll soon have to re-enable FGRPB or run out of CPU work all togheter
Christian Beer wrote:You need
)
I noticed an entry in the server log (for Arecibo GPUs, but I think it was generic): "only one Beta task per workunit". If that's in effect, and there are no non-Beta apps yet, then the WUs will never have a validation partner.
Richard Haselgrove
)
That is only the case when there is a non-beta app_version and a beta app_version (as is the case for BRP4G). The Beta flag is set for all GW app_versions so they will validate against each other (once I test and activate the validator).
There seems to be something wrong with the scheduler which I will investigate once I fixed the problem with the datafiles.
I had a hard time getting the
)
I had a hard time getting the beta too. I had to request it three or four times, even though I had four empty cores and nothing else running. Finally, I picked up one work unit, and then several more followed on subsequent requests. I ended up with 18, of which 14 completed without errors and the other 4 errored out (BOINC 7.6.33, Win7 64-bit, i7-4771).
I have had Test Applications
)
I have had Test Applications selected all the time, I dis-abled Gamma-Ray work to try and get the new work units.
It seems that something has worked, I managed to get 4 Multi-Directed work units a couple of hours ago.
Only on my Windows machine so far but I will wait.
A curious thing that I have noticed is that the Gamma-Ray work units state they are "Beta" in their name, but the Multi-Directed work units do not state they are Beta so is the tuning done and these are production work units?
Conan
I just aborted about 300 GW
)
I just aborted about 300 GW as there was no way for my machines (i7s 6950X, 6900, 5960X all OC to 4.0GHz) to process them all in less than 72 hours. I'm finding that if all CPU cores are engaged that work is taking anywhere from 9-11 hours for the longest work units. Haven't seen any validations yet.
I removed the Beta flag on
)
I removed the Beta flag on the applications. Everyone should get some GW work now. I'm still monitoring the scheduler and may start to take a look at the results waiting for validation.
I'll be watching the project over the weekend from time to time in case something else breaks.
The datafile issue is now understood and only affects a small range of tasks which I'm going to cancel once they show symptoms.
Of the seven tasks sent to
)
Of the seven tasks sent to one of my hosts within a few seconds of each other about 12 Oct 2016, 22:42:30 UTC, it remains true that all seven show the second task as unsent at this time. For example this one.
On another host, a single one of these tasks did get the second task sent to a partner, who returned his a while after me, but still over half a day ago, at 14 Oct 2016, 2:21:52 UTC. This WU is perhaps waiting for a manually started run of the validator, as I suppose that is not running automatically just yet.
Now I have got LOTS of GW
)
Now I have got LOTS of GW units. Very nice.
Unfortunately they have a 3 day deadline and given that they take a long time to complete I fear only a few of them will be done in time.