No idea where the extra 3 hours come from on a quicker machine.
Found the reason for that. The CPU-usage of both GPU tasks is taking away the CPU time of the FGRP. Disabling GPU brings the time back to what i would expect. Guess i demand too much :)
Got quite a lot of download errors for the new FGRP3 file.
By "new FGRP3 file" do you mean the next data file in the series - LATeah0057C.dat perhaps?
Quote:
Sending of FGRP3 tasks have been turned off during investigation.
I notice that there is no longer any column for FGRP3 in the "Workunits and Tasks" section on the status page. Perhaps this is what happens if you 'turn off' tasks?
Is the FGRP3 problem related to what happened with S6CasA? Is FGRP3 likely to be out for the whole weekend?
By "new FGRP3 file" do you mean the next data file in the series - LATeah0057C.dat perhaps?
Originally I meant this. Turend out, however, that it's actually the new application binary.
Quote:
Sending of FGRP3 tasks have been turned off during investigation.
I notice that there is no longer any column for FGRP3 in the "Workunits and Tasks" section on the status page. Perhaps this is what happens if you 'turn off' tasks?
Yes. We "deprecated" the whole application.
Quote:
Is the FGRP3 problem related to what happened with S6CasA?
Apparently yes. The problem seems to be the download mirrors at UWM, which serve the GW data and all application binaries.
Quote:
Is FGRP3 likely to be out for the whole weekend?
I'm working to avoid this. Currently I "turned off" the GW work again to take some load off the mirrors, when these relax a bit I'll try to switch FGRP3 on again.
Be warned that it looks like SETI failed again about an hour ago - and midnight Friday local time isn't a good time to expect a quick fix.
It's the sort of failure which will probably ramp up the demand for BRP Arecibo data here gradually - and most people crossing over will have their binaries already, so UWM shouldn't get too much of a hit.
I guess that couple of things coincided here to put us in a bit of trouble:
- technical issues with the UWM download mirrors (originally there were five machines there, two of which already died, the others crawl along but apparently aren't very performant)
- the announcement of the HTC - BOINC partnership at MWC on Monday, which brought quite a number of new participants with Android devices (not much computing power, but all downloading applications)
- a new application version for the (apparently popular) FGRP3 search
- SETI down again (sic)
Anyway, shifting load back and forth between the mirrors (and disabling GW work) apparently improved the situation far enough that I dare to turn FGRP3 back on again.
GW work will remain turned off probably until I'm back in the office on Monday.
- the announcement of the HTC - BOINC partnership at MWC on Monday, which brought quite a number of new participants with Android devices (not much computing power, but all downloading applications)
Anyway, shifting load back and forth between the mirrors (and disabling GW work) apparently improved the situation far enough that I dare to turn FGRP3 back on again.
GW work will remain turned off probably until I'm back in the office on Monday.
RE: No idea where the extra
)
Found the reason for that. The CPU-usage of both GPU tasks is taking away the CPU time of the FGRP. Disabling GPU brings the time back to what i would expect. Guess i demand too much :)
Hallo! The difficulty is,
)
Hallo!
The difficulty is, there are no other purly CPU tasks available at now.
Kind regrards and happy crunching
Martin
Got quite a lot of download
)
Got quite a lot of download errors for the new FGRP3 file.
Sending of FGRP3 tasks have been turned off during investigation.
BM
BM
RE: Got quite a lot of
)
By "new FGRP3 file" do you mean the next data file in the series - LATeah0057C.dat perhaps?
I notice that there is no longer any column for FGRP3 in the "Workunits and Tasks" section on the status page. Perhaps this is what happens if you 'turn off' tasks?
Is the FGRP3 problem related to what happened with S6CasA? Is FGRP3 likely to be out for the whole weekend?
Cheers,
Gary.
Hallo Gary! I got my first
)
Hallo Gary!
I got my first S6CasA-tasks at 28 Feb 2014 16:52:46 UTC and they deliver constantly since than.
Kind Regards and happy crunching
Martin
RE: RE: By "new FGRP3
)
Yes. We "deprecated" the whole application.
Apparently yes. The problem seems to be the download mirrors at UWM, which serve the GW data and all application binaries.
I'm working to avoid this. Currently I "turned off" the GW work again to take some load off the mirrors, when these relax a bit I'll try to switch FGRP3 on again.
BM
BM
Be warned that it looks like
)
Be warned that it looks like SETI failed again about an hour ago - and midnight Friday local time isn't a good time to expect a quick fix.
It's the sort of failure which will probably ramp up the demand for BRP Arecibo data here gradually - and most people crossing over will have their binaries already, so UWM shouldn't get too much of a hit.
Thanks for the note. I
)
Thanks for the note.
I guess that couple of things coincided here to put us in a bit of trouble:
- technical issues with the UWM download mirrors (originally there were five machines there, two of which already died, the others crawl along but apparently aren't very performant)
- the announcement of the HTC - BOINC partnership at MWC on Monday, which brought quite a number of new participants with Android devices (not much computing power, but all downloading applications)
- a new application version for the (apparently popular) FGRP3 search
- SETI down again (sic)
Anyway, shifting load back and forth between the mirrors (and disabling GW work) apparently improved the situation far enough that I dare to turn FGRP3 back on again.
GW work will remain turned off probably until I'm back in the office on Monday.
BM
BM
RE: - the announcement of
)
Reference that, see my question here
Claggy
RE: Anyway, shifting load
)
We will be ready and waiting as always Bernd
Have a fine weekend