| Author |
Message |
|---|
|
|
|
Just returned a load of PPS LLR tasks which validated fine but granted zero credit. Example Happening to other hosts too. Server problem?
____________
  |
|
|
RytisVolunteer moderator
Project administrator
Send message
Joined: 22 Jun 05
Posts: 2653
ID: 1
Credit: 109,689,826
RAC: 48,854
                     
|
|
I think it's OK now: I upgraded the server and it stopped storing claimed credit! I fixed the problem, but there doesn't seem to be a way to extract credit for the results that were not updated correctly...
[update] We will do our best to grant at least approoximate credit for the work done though.
____________
|
|
|
Ken_g6 Volunteer developer
Send message
Joined: 4 Jul 06
Posts: 940
ID: 3110
Credit: 265,196,425
RAC: 105,920
                           
|
|
Hey, I can't sort my results by application anymore! Does this have something to do with your "upgrade"?
Edit: I guess no one can see that link, but I can't sort my computer's results that way either.
____________
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
             
|
Hey, I can't sort my results by application anymore! Does this have something to do with your "upgrade"?
Edit: I guess no one can see that link, but I can't sort my computer's results that way either.
Yes, result of server upgrade. :( They will return...hopefully before the Challenge. :)
____________
|
|
|
|
|
|
DA and his gang at their best.. :( |
|
|
|
|
DA and his gang at their best.. :(
This is going to be interesting... i do hope that the <primary_platform_only> flag is set on the server to avoid sending 32 bit apps to 64 bit hosts, to avoid the mess that happened over at aqua where this stupidity wasted a lot of cpu time....
|
|
|
RytisVolunteer moderator
Project administrator
Send message
Joined: 22 Jun 05
Posts: 2653
ID: 1
Credit: 109,689,826
RAC: 48,854
|
This is going to be interesting... i do hope that the <primary_platform_only> flag is set on the server to avoid sending 32 bit apps to 64 bit hosts
It is not, because LLR app is 32bit only.
____________
|
|
|
|
|
This is going to be interesting... i do hope that the <primary_platform_only> flag is set on the server to avoid sending 32 bit apps to 64 bit hosts
It is not, because LLR app is 32bit only.
So i need to build myself an app_info.xml for the 321 sieve challenge to avoid getting the 32 bit app....
|
|
|
 HAmstyVolunteer tester
Send message
Joined: 26 Dec 08
Posts: 132
ID: 33421
Credit: 12,510,712
RAC: 0
       
|
|
no, this challange is about sieving and your 64bit windows host will get the 64bit windows app
____________
 |
|
|
|
|
no, this challange is about sieving and your 64bit windows host will get the 64bit windows app
as long as the <primary_platform_only> flag is not set server side you cannot avoid getting the 32 bit sieve app. This applies to all 64 bit linux,win and mac. This is supposed to be a feature of the new server code to measure wich app works best....
|
|
|
RytisVolunteer moderator
Project administrator
Send message
Joined: 22 Jun 05
Posts: 2653
ID: 1
Credit: 109,689,826
RAC: 48,854
|
|
That's a new one. We're evaluating our options, but it might be too late for us to make a change this close to the challenge (we basically need to cheat with LLR apps and tell BOINC we also have 64bit version of them, and then use primary_platform_only).
____________
|
|
|
|
|
That's a new one. We're evaluating our options, but it might be too late for us to make a change this close to the challenge (we basically need to cheat with LLR apps and tell BOINC we also have 64bit version of them, and then use primary_platform_only).
Yes. for LLR that's the only option to rename the 32 bit binary and fake a 64 bit one.
Another problem of that new feature is that on 64 bit linux, there's no way to tell if the 32 bit compat libs are installed. So you end up getting a whole bunch of trashed tasks on 64 bit only hosts.
EDIT
does anyone here have a app_info.xml for 321 sieve as a template ??? |
|
|
|
|
That's a new one. We're evaluating our options, but it might be too late for us to make a change this close to the challenge (we basically need to cheat with LLR apps and tell BOINC we also have 64bit version of them, and then use primary_platform_only).
that's probably the only way to quirk around that really bright concept.
|
|
|
RytisVolunteer moderator
Project administrator
Send message
Joined: 22 Jun 05
Posts: 2653
ID: 1
Credit: 109,689,826
RAC: 48,854
|
|
I compiled a custom version of the server, disabling this "feature" (actually, I made it to accept 0 results to determine speed; may or may not work - I might have to switch to 1 result). Please test if it really sends out correct bitness work.
____________
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
as long as the <primary_platform_only> flag is not set server side you cannot avoid getting the 32 bit sieve app. This applies to all 64 bit linux,win and mac. This is supposed to be a feature of the new server code to measure wich app works best....
NOTE: this may also be the reason for the CUDA errors for 1.27...32 bit going to 64 bit. We'll have to investigate that after this is solved here. :( If it gets fixed here, we hope it fixes the CUDA issues as well.
____________
|
|
|
|
|
NOTE: this may also be the reason for the CUDA errors for 1.27...32 bit going to 64 bit. We'll have to investigate that after this is solved here. :( If it gets fixed here, we hope it fixes the CUDA issues as well.
i have run some WU's on a 64-bit windows host during the last hours (CPU and GPU) and did not see any 32-bit apps yet.
|
|
|
|
|
NOTE: this may also be the reason for the CUDA errors for 1.27...32 bit going to 64 bit. We'll have to investigate that after this is solved here. :( If it gets fixed here, we hope it fixes the CUDA issues as well.
i have run some WU's on a 64-bit windows host during the last hours (CPU and GPU) and did not see any 32-bit apps yet.
The problem is that this feature seems to pick hosts at random to test apps. So there's no real way to trigger or test if Rytis custom version actually fixed it. The only way is to monitor your host.
FWIW, Macs seem to be most affected by it(only had it happening here at one win64 hosts). At least that's what i've seen over at AQUA. |
|
|
Ken_g6Volunteer developer
Send message
Joined: 4 Jul 06
Posts: 940
ID: 3110
Credit: 265,196,425
RAC: 105,920
|
|
So does that mean that those who followed the advice in the race thread and ran a few WUs a week ago won't have any problem?
____________
|
|
|
RytisVolunteer moderator
Project administrator
Send message
Joined: 22 Jun 05
Posts: 2653
ID: 1
Credit: 109,689,826
RAC: 48,854
|
|
If my fix worked, nobody will have problems; if it didn't, running work a week ago does not have any impact since the server upgrade was done two days ago.
____________
|
|
|
Sysadm@NbgVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1233
ID: 18646
Credit: 918,613,436
RAC: 389,671
           
|
|
my 64bit Linux crunching with the x86_64-app
so I think, your fix works for me ...
its challenge time - keep on crunching
____________
Sysadm@Nbg
my current lucky number: 113856050^65536 + 1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
 |
|
|
|
|
I compiled a custom version of the server, disabling this "feature" (actually, I made it to accept 0 results to determine speed; may or may not work - I might have to switch to 1 result). Please test if it really sends out correct bitness work.
I've been crunching the Challenge for 7 hours and I've seen no cross-platform 321 sieve apps ;D
____________
  |
|
|
