While looking at some ATI results, I've come across this host,
5292089, making only GPU (9800GTX)errors.
CPU results are OK.
This host shouldn't get so much tasks, CPU only f.i., IMHO.

---

Wow, that's quite ridiculous.

Resends-alert?

Unfortunately it is not producing only errors, about 1 in 25 are being reported by the CUDA apps as successful, and thereby boosting the quota back up. Later they're being found invalid, but that doesn't help much.

The anonymous user of course can't be contacted, and the 9800 GTX+ has enough memory that both stock 6.08 and 6.09 are being used and have their separate quotas, which sort of doubles the problem.

In the big picture with ~60000 results being reported each hour, one host like that isn't much of a problem, but it is certainly reducing the overall productivity.

                                                                   Joe

Don't look at my errors :(

I managed to reset the SETI project during a fit of Boinc lag last week, borking 5K tasks.

---

Executive Director GPU Users Group Inc. -
brad@gpuug.org

OOoo, you, you, badly borked your BOINC.
Sounds nasty,
Can you get tablets to help with that :¬)

My pending list is going through the roof again but when you get teamed with these Anonymous users with faulty/unserviced or poorly setup machines that cut large chunks of "inconclusives" through my list its bad,

a few examples,

3378825
6469701
6256705
6589662

but when they double up on you its even worse then throw in a good mix of people using the sleep bug version CUDA drivers (I sent off a lot of PM's today but how many actually get looked at will be another matter) and things get really crazy.

Cheers.

---

Someone on the server side of our project should add some measures, to protect us against people like that. I.e. an error cap, if you produce this many errors, you get cut off of whatever work you are erroring on. Then the person can come here and make themselves known in order to get a "talkin' to" and get their s*** straight.

---

#resist

Unfortunately it is not producing only errors, about 1 in 25 are being reported by the CUDA apps as successful, and thereby boosting the quota back up.

That's what I always ask myself, why is the quota not increased by 1 (or 2 at most) for each successful result just like it is decreased by 1 for each error? That would give a GPU 8 new tasks, that should be enough to proof if it's reliable again. I don't think it's not too much to expect, that a host should return at least 50% of his tasks without an error (actually I'd expect at least 95% to be "successful" for a properly working machine, but a machine that has been fixed must have a chance to build up the quota in reasonable time).

---

Don't look at my errors :(

I managed to reset the SETI project during a fit of Boinc lag last week, borking 5K tasks.

did that a couple of weeks ago myself ... ~3000 down the tube ... maybe we should have a borking sticky ...

---

Don't look at my errors :(

I managed to reset the SETI project during a fit of Boinc lag last week, borking 5K tasks.

did that a couple of weeks ago myself ... ~3000 down the tube ... maybe we should have a borking sticky ...

Those that aborted work I never sent a PM to as I figured that those who did found their problem just like those who's video cards had disappeared from their setups but this week I've found a lot of bad wingmen (and I mean a great deal but some others had not yet had enough proof of a problem though that also may change next week).

BTW out of about 24 PM's that I sent off around 12hrs ago I've only received 1 reply so far but maybe that will increase by another in another 12hrs (but I may name & shame those who havn't done any corrective action by next week).

Cheers.

---

Yep, I noticed that in my pending list over time it contains a wingman called Anonymous, in inconsistants.
I also noticed that Anonymous has a number of host computers associated with this name.

One of which is 3378825, which at the time of this reply had

12795 Pending
88 Valid
1144 Invalid
5 Error

That's one awful lot of Invalids , no wonder we got to rehash the work to get valid results.

I also noticed that Anonymous has a number of host computers associated with this name.

Each and every user that has his/her computer(s) hidden is represented as 'Anonymous'!

Gruß,
Gundolf

Yep, I know that

so is there really no way to remedy this situation? Can't we email or message the administrators at S@H and report this, and perhaps they can take a few moments to figure out a permanent fix for this (like the error quota mentioned below)?

From previous similar discussions - wasn't there some sort of a 'blacklist' capability within BOINC? Can't remember whether it operated at the host or account level, but I agree - something ought to be done about hosts/users who just spit out rubbish and squander resources!

---

From previous similar discussions - wasn't there some sort of a 'blacklist' capability within BOINC? Can't remember whether it operated at the host or account level, but I agree - something ought to be done about hosts/users who just spit out rubbish and squander resources!

We actually have something, that could be able to automatically stop such hosts: quota system. Just the current settings are making it pretty useless, I see that here, I see that on Milkyway and on other projects it won't be much different.

---

From previous similar discussions - wasn't there some sort of a 'blacklist' capability within BOINC? Can't remember whether it operated at the host or account level, but I agree - something ought to be done about hosts/users who just spit out rubbish and squander resources!

http://boinc.berkeley.edu/trac/wiki/BlackList#

so is there really no way to remedy this situation? Can't we email or message the administrators at S@H and report this, and perhaps they can take a few moments to figure out a permanent fix for this (like the error quota mentioned below)?

The S@H project can use whatever features BOINC provides, although BOINC is LGPL and they could modify it the time and effort involved would be far more than "a few moments".

There's a <daily_result_quota> project option which is currently at 100. If it were practical to reduce it to 20 or so, the problem would be much reduced. However, the CUDA apps have some builtin "unsupported function" -12 errors which are no fault of the owner or host and occur often enough that a lower base for the quota would impact the productivity of some of the best systems. I can't really judge whether something like a 50 setting might be practical or not, but my guess is not.

There have been several discussions on the boinc_dev mailing list related to the quota logic, but Dr. Anderson has not been persuaded that any major change is needed. If enough projects complained, that could change; the server code is very much intended to target project needs rather than what we external participants see as problematic.

                                                                  Joe

Host
6469701 has a lot of overflows or -9 errors.
Yet another anonymous host with an GTX 560 and a GTS8800!

---