Message boards :
Number crunching :
Host with strange error
Message board moderation
Author | Message |
---|---|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Many times restarted @100%, the same counts every WU, too many credit claimed for too small computation time for such CPU. http://setiathome.berkeley.edu/results.php?hostid=2046662&offset=120 Resuts are not validated, but... Maximum daily WU quota per CPU 100/day Why quota doesn't go down ???? |
kittyman Send message Joined: 9 Jul 00 Posts: 51469 Credit: 1,018,363,574 RAC: 1,004 |
Whatever the problem is, the host is not being granted any credit for the work.... "Freedom is just Chaos, with better lighting." Alan Dean Foster |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Yes, but quota still 100 WU per day - that is the problem (it leech WUs at 17sec per WU rate! ) |
UncleVom Send message Joined: 25 Dec 99 Posts: 123 Credit: 5,734,294 RAC: 0 |
Yes, but quota still 100 WU per day - that is the problem (it leech WUs at 17sec per WU rate! ) My guess a virtual machine gone wild. Why it is not stopped from getting more work may be related to this. IIRC the work unit quota is 100 per CPU core. UncleVom |
gomeyer Send message Joined: 21 May 99 Posts: 488 Credit: 50,370,425 RAC: 0 |
I think the quota only goes down for errors. These are marked as "Success" even tho' he is indeed returning bad results. It looks like he got 100 for June 22, then nothing for four hours since, so he seems to have hit the quota. Glad he is not running a quad! |
OzzFan Send message Joined: 9 Apr 02 Posts: 15691 Credit: 84,761,841 RAC: 28 |
IIRC the work unit quota is 100 per CPU core. It is 100 per core, but I believe BOINC still treats the number as a whole, per system. So if a single core returns a bad result, it will be lowered to 99 per core, even on a quad core. As for a possible answer to the question "How is this person allowed to download so much while returning bad results?" Could it be a bad overclock returning incorrect results instead of invalid results? I think the quota only covers a CPU returning invalid results but does nothing against incorrect results. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
I think the quota only goes down for errors. These are marked as "Success" even tho' he is indeed returning bad results. That's right. Whatever has gone wrong with the host began on June 17, and the 5.8.16 core client doesn't recognize there's a problem. I think the stderr.txt stuff indicates the core client hasn't been able to clear out the slot, so the app is running WUs starting from the leftover final checkpoint. More recent versions of BOINC do checking and create a new slot if they were unable to clear an old one. Joe |
OzzFan Send message Joined: 9 Apr 02 Posts: 15691 Credit: 84,761,841 RAC: 28 |
I think the quota only goes down for errors. These are marked as "Success" even tho' he is indeed returning bad results. I thought "success" merely meant that the WU completed and was successfully returned to the server, not necessarily meaning a 'good' or 'bad' result. |
Jakob Creutzfeld Send message Joined: 13 Oct 00 Posts: 611 Credit: 2,025,000 RAC: 0 |
Seems like BOINC is not able to clean up the working slot directory, so each unit uses an old result/state file, "restarts" at 99.78%, take only some seconds to completion, and claims too much credit (AFAIK credits are managed by BOINC, not the science app). Something like wrong/insufficious access/directory rights maybe? Andy |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
So, if I understood right, quota mechanism leaves BOINC unprotected in case of intentional modification of science app to return bad results marked as success ones? It could lead to some kind of DoS attack with great network load (new WU could be downloaded at very high rate in such situation). |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
So, if I understood right, quota mechanism leaves BOINC unprotected in case of intentional modification of science app to return bad results marked as success ones? I doubt the difference between a quota of 100 and a reduced quota of 1 per day would make much difference to someone designing that sort of DoS attack, they'd need a large number of hosts in either case to make the attack effective. I just sent a polite PM to the system's owner, though as there are no forum posts listed for the account he probably won't see it. Joe |
Fred J. Verster Send message Joined: 21 Apr 04 Posts: 3252 Credit: 31,903,643 RAC: 0 |
So, if I understood right, quota mechanism leaves BOINC unprotected in case of intentional modification of science app to return bad results marked as success ones? But, why would someone turn in faulty results, in the first place. The credit claimes are all the same and kind off high! Stupid question maybe, how could one 'proof', f.i. this is fixed? |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.