Weird quorum behaviour
16 Aug 2006 13:04:45 UTC
Topic 191687
(moderation:
This workunit shows a weird quorum behaviour, in my opinion : http://einsteinathome.org/workunit/11407981
A third result id was requested probably because the first two results were different (what else could it be?) but when the third result gets in it is rejected and don't get credits, and the first two are now valid. If the first results were valid, why ask for a third ?
Weird quorum behaviour
)
I didn't look carefully at the time reported column, now I understand better what happened.
The first result was late (due time Aug 10 reported Aug 11), so a third result was asked. But when the first result came in, late, it was valid and got credits... wasting the third computer time and not giving it any credit.
I understand better, but I don't think this is fair.
It could truely be that the
)
It could truely be that the result was invalid. Look at past history on that 3rd computer, it has previous bad results. Out of 15 results returned and validated, 5 were bad. The 4 previous he was one of the first 2 to send the result back in.
He is using a 3rd party BOINC product that tries to adjust speeds and benchmarks, which no longer matter in computing on Eisntein.
Even if the result is
)
Even if the result is invalid, it does not make sense to ask for a third advice when quorum has been agreed. It's also normal to be one of the first 2 computers to return a result if it is invalid... else there would be no need to ask a third computer and you can't know the result is invalid if it's not reported yet...
Btw, the third computer is mine and I replaced a faulty piece of RAM last week. Since then, there was no invalid result and there has never been any before, AFAIK.
For the boinc client, yes it's Truxoft calibrating client but as you said, that doesn't matter anymore and I don't feel like reinstalling boinc on that box.
RE: I didn't look carefully
)
the third result would have been granted credits if it were valid!
Why it has been set invalid I don't know.
Udo
Udo
I don't see anything
)
I don't see anything suspicious in there.
The first two Tasks have been sent out on Jul 26, which means the deadline was Aug 10. At that date only one of them has been reported, so a third Task was sent out.
While that was in progress, the overdue first Task was reported. The two tasks formed a quorum and have been found to agree and thus to be valid. When the third task was reported, it was found not to agree with the canonical result (for whatever reason) and thus was marked invalid and granted no credit.
The reason for the third task beeing invalid is definitely not that it came late (else it would have been marked "too late").
BM
BM