Current Problem with Collatz [info]

Good at maths? Then why not figure on helping these projects - BURP, Citizen Science Grid, Collatz, Enigma, Gerasim, Moo! Wrapper, NFS, NumberFields, PrimeGrid, WEP+M2.
Post Reply
chriscambridge
Active UBT Contributor 1+ yr
Posts: 1760
Joined: Mon Aug 08, 2016 1:56 pm
Location: Duroliponte | Cantabrigia

Current Problem with Collatz [info]

Post by chriscambridge » Mon Apr 02, 2018 1:14 pm

Just so everyone is aware an issue has been raised with Collatz, with some hosts completing tasks in around 5 seconds.

via: https://www.reddit.com/r/gridcoin/comme ... atz_hosts/

Response from Project Admin:
Project Down Until...

It seems a few people have come up with a way to game the system.
Until I figure out what damage they have done and how many thousands
of work units need to be re-issued, the daemons are shut down. You
should still be able to return any completed work units and they will
eventually be validated and credit given if valid, but no new work units
will be sent until new applications and server daemons have been developed.


Image
Image

UBT - Timbo
UBT Forum Admin
Posts: 7655
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Re: Current Problem with Collatz [info]

Post by UBT - Timbo » Mon Apr 02, 2018 2:19 pm

Hi Chris

Thanks for the heads up.

One wonders WHY people do this...(yes I know they do it just for the credits)...but the project suffers if the work returned isn't "kosher" and as noted the original tasks have to be returned to the "available work" pile so that people crunching this project, for the projects benefit, can return a real result.

I really hope the admins of Collatz ban the perpetrators and publish who they are.

AND I hope they nullify any credits earned and we can then see which teams suddenly lose lots of credits - I really hope none of the UBT membership are involved.

regards
Tim

PS: The computer shown above is "crunching" !!

https://boinc.thesonntags.com/collatz/r ... e=4&appid=

In the last DAY this computer has "earned" 925,120,000 Collatz credits - that should make it easy to spot on BOINCstats !! Plus it has probably earned lots of GRC and Etherium tokens too !!

chriscambridge
Active UBT Contributor 1+ yr
Posts: 1760
Joined: Mon Aug 08, 2016 1:56 pm
Location: Duroliponte | Cantabrigia

Re: Current Problem with Collatz [info]

Post by chriscambridge » Mon Apr 02, 2018 2:49 pm

I totally agree.

In the above case, the host was a GRCPool member - however their account has now been blocked and none of the GRC will be issued to them.

So they have not only wasted everyone else's time, but also their own. Like you mentioned, this kind of issue is pretty easy to spot so I am not sure why they really bothered.

I think the most that could maybe be published is their IP and perhaps a GRC Wallet address; which may or may not be useful in tracking down who this/they are.
Image

UBT - Timbo
UBT Forum Admin
Posts: 7655
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Re: Current Problem with Collataz [info]

Post by UBT - Timbo » Mon Apr 02, 2018 7:06 pm

Hi Chris

It could be this/these "user(s)", grcpool.com-3, sitting at the top of the table":

https://boincstats.com/en/stats/86/user ... /bestxdays

and with these recent stats:

https://boincstats.com/en/stats/86/user/list/10/0/0

and these are the "hosts" earning the credits:

https://boincstats.com/en/stats/86/host ... /0/1584175

two of which seem to be earning the most !! (and both have same CPU's)
Intel(R) Xeon(R) CPU E3-1230 v3 @ 3.30GHz 4(8) Windows 7 Professional x64 Edition, Service Pack 1, (06.01.7601.00) 925,120,000 925,120,000
Intel(R) Xeon(R) CPU E3-1230 v3 @ 3.30GHz 4(8) Windows 7 Professional x64 Edition, Service Pack 1, (06.01.7601.00) 922,240,000 922,240,000
I am purely GUESSING at the above as this seems to tie in with a user who has generated approx 900 million credits per day - in fact over the last couple of days it's been more like 1 billion credits per day...

regards
Tim

chriscambridge
Active UBT Contributor 1+ yr
Posts: 1760
Joined: Mon Aug 08, 2016 1:56 pm
Location: Duroliponte | Cantabrigia

Re: Current Problem with Collatz [info]

Post by chriscambridge » Mon Apr 02, 2018 8:12 pm

I wouldn't worry too much as GRCPool admin is already on the case.
Image

UBT - Timbo
UBT Forum Admin
Posts: 7655
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Re: Current Problem with Collatz [info]

Post by UBT - Timbo » Tue Apr 03, 2018 1:31 pm

chriscambridge wrote:
Mon Apr 02, 2018 8:12 pm
I wouldn't worry too much as GRCPool admin is already on the case.
I'm not so worried about the GRC - more the problem is that the Collatz admins must now spend time on checking all the results from this scammer, nullifying them and then re-writing the app code to prevent this happening again.

So, thousands of Collatz crunchers will not get any new work for a time, until it is fixed. Plus the admins will need to spend time and money on developing the new app.

Maybe GRC can (should?) pay them for their time? as it is one of their "pool" users who has done this.

I hope they can get their IP address and block them.

regards
Tim

Jeffers
Active UBT Contributor 15+ yrs
Posts: 1577
Joined: Mon Jul 24, 2006 1:00 am
Location: Halifax, West Yorks.

Re: Current Problem with Collatz [info]

Post by Jeffers » Tue Apr 03, 2018 2:11 pm

I've been running a lot of Collatz and I've still got quite a few tasks waiting to run. Should I just leave everything as it is and wait for things to get sorted? I've re-enabled Primegrid to keep stuff available for my GPU when Collatz runs out.
Image

UBT - Timbo
UBT Forum Admin
Posts: 7655
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Re: Current Problem with Collatz [info]

Post by UBT - Timbo » Tue Apr 03, 2018 2:37 pm

Hi Jeff

Let them run....there's no issue with tasks sent out to genuine crunchers.

The issue will be that the Collatz admins will stop generating new work to be crunched, until they have re-developed the Collatz app that does the "donkey work".

But feel free to also allow tasks from other GPU projects such as PrimeGrid and Moo!

When a new Collatz app is available, you may be asked to delete the project from BOINC Manager and to then add it back again, as this will "force" a download of the new app. Also, make a back-up copy of your "app_config" file (in your Collatz directory) as you may find that gets deleted if you remove the project from BOINC Manager.

(It is possible that Collatz will automatically update the app, once they have fixed it...but that won't be known until they sort it out).

regards
Tim

Woodles
Active UBT Contributor 20+ yrs
Posts: 10191
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: Current Problem with Collatz [info]

Post by Woodles » Tue Apr 03, 2018 3:08 pm

Hi Jeff,

Any work already sent out can be safely crunched and returned. there won't be any new work. To be ultra safe, you can pause any work that isn't near it's deadline yet just in case.

The problem is that until the admin has decided what work has actually been carried out then nothing returned is going to get validated so do the work, return the task and sometime in the future you'll get the credit for it.

I would imagine it's going to take a few days to sort out so you'll need a different project unless you have a LOT of work in your cache.

Mark

Woodles
Active UBT Contributor 20+ yrs
Posts: 10191
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: Current Problem with Collatz [info]

Post by Woodles » Tue Apr 03, 2018 3:15 pm

Hi Chris,

My understanding was that the owner of the hosts wouldn't be getting paid but their credit was still being counted? Doesn't this mean that all other members of Gridcoin crunching Collatz will be getting a relatively smaller payment than they should?

Part of a post from Brian:
From a pool perspective, the account responsible for the hosts is locked and will not payout anymore GRC into the owners Gridcoin wallet. The Gridcoin network and pool will continue to recognize the host credit until Collatz eliminates the statistics.
Mark

chriscambridge
Active UBT Contributor 1+ yr
Posts: 1760
Joined: Mon Aug 08, 2016 1:56 pm
Location: Duroliponte | Cantabrigia

Re: Current Problem with Collatz [info]

Post by chriscambridge » Wed Apr 04, 2018 9:39 pm

HI Mark,

What will happen is the host will still be allocated GRC due to RAC/MAG, however their account has now been frozen and no GRC will be issued to that wallet; indeed I am sure reverse transaction analysis is being done on that wallet address right now to try and find out who owns it.

Additionally I am guessing the IP, Mac address (of all related hosts), Wallet address (and related wallet addresses) will now be block/banned across Gridcoin.

Because of the massive numbers this host has done, yes it does mean everyone else who has crunched Collatz under Gridcoin will/has lost out quite substantially.

In terms of the fraudulent GRC; I guess there will probably be a GRCPool vote to determine what should happen to the frozen GRC. Either it will be "rained down" on other Collatz crunchers (as they are the ones who have lost out), or perhaps even sent to the admin of Collatz as a donation.
Image

UBT - Timbo
UBT Forum Admin
Posts: 7655
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Re: Current Problem with Collatz [info]

Post by UBT - Timbo » Sat Apr 21, 2018 12:23 am

Hi all

I'd not heard much about the ongoing "fix" for this issue, so I took a look at the Collatz website.

It looks like their entire message board has been re-installed and all the previous messages (many years worth) have GONE !!

https://boinc.thesonntags.com/collatz/forum_index.php

It also looks like some the historical data on GPUs has vanished too (mainly Mac and Linux OS) !! (CPU info still seems to be there though).

https://boinc.thesonntags.com/collatz/gpu_list.php and

Also, due to the changes on the website, one needs to REMOVE the project from BOINC Manager and then select it from the "drop down" list and add it back again.

Needless to say *if* you are using an optimised CONFIG.XML file, make a copy of this FIRST, BEFORE you use BOINC Manager to remove the project, otherwise you may well lose said file(s).

If you do lose the file, then someone has already posted some info on the various options available for various GPUs, here:

https://boinc.thesonntags.com/collatz/f ... hp?id=8#69

So, be careful.

Hopefully, new Collatz tasks will be forthcoming soon...

regards
Tim

bigsinky
Active UBT Contributor 15+ yrs
Posts: 270
Joined: Wed Mar 28, 2018 8:27 pm
Location: Northern Ireland
Contact:

Re: Current Problem with Collatz [info]

Post by bigsinky » Fri Dec 06, 2019 8:46 am

whats the craic with these compute WUs. either 4 seconds long or 20 minutes long. its f**king may Collatz RAC up no end.
Image

Woodles
Active UBT Contributor 20+ yrs
Posts: 10191
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: Current Problem with Collatz [info]

Post by Woodles » Fri Dec 06, 2019 9:55 am

Hi Neil,

Your 4 second ones:
worker: error reading input file.
Error -102. Processing Aborted.
Bad workunit, no input file - https://boinc.thesonntags.com/collatz/f ... =2110#2110
Nothing you can do, once enough people have failed them then they'll get removed from the system, until then, we have to put up with it or change projects.

Your twenty minute one:
<stderr_txt>
Collatz Conjecture Sieve 1.30 Windows x86_64 for OpenCL
Written by Slicker (Jon Sonntag) of team SETI.USA
Based on the AMD Brook+ kernels by Gipsel of team Planet 3DNow!
Sieve code and OpenCL optimization provided by Sosiris of team BOINC@Taiwan
Collatz Config Settings:
verbose 1 (yes)
kernels/reduction 48
threads 2^8 (256)
lut_size 17 (1048576 bytes)
sieve_size 2^30 (51085096 bytes)
sleep 1
cache_sieve 1 (yes)
reducecpu 0 (no)
Processor Type NVIDIA
Collatz Conjecture Sieve 1.30 Windows x86_64 for OpenCL
Written by Slicker (Jon Sonntag) of team SETI.USA
Based on the AMD Brook+ kernels by Gipsel of team Planet 3DNow!
Sieve code and OpenCL optimization provided by Sosiris of team BOINC@Taiwan
Collatz Config Settings:
verbose 1 (yes)
kernels/reduction 48
threads 2^8 (256)
lut_size 17 (1048576 bytes)
sieve_size 2^30 (51085096 bytes)
sleep 1
cache_sieve 1 (yes)
reducecpu 0 (no)
Processor Type NVIDIA


Unhandled Exception Detected...
Seems to have restarted part way through the initialisation and got hung up.
Might be worth backing off the clocks a little or reducing the number of threads in the config file (although they're acceptable settings) .... or since it appears to have been a one off, probably just a noise spike or glitch and it won't happen again.

[/edit] Has Windows updated recently? It always screws up my graphics drivers when it does and causes errors until I reinstall them from Nvidia?

Mark

bigsinky
Active UBT Contributor 15+ yrs
Posts: 270
Joined: Wed Mar 28, 2018 8:27 pm
Location: Northern Ireland
Contact:

Re: Current Problem with Collatz [info]

Post by bigsinky » Sat Dec 07, 2019 1:10 am

mark,

thanks for looking into this. i am fastidious when it comes to drivers. Updated BIOS last week and the the new Nvdia driver always downloads as i have the GeForce Experience app check once a day for new Nvidia drivers so they are always up to date. the gpu_config.xml annd app_config.xml files for collatz have worked flawlessly up to a few weeks ago. i passed them to damien and his RAC has shot up significantly. I am back to about 20 million a day which is about right for 2 x 1080ti. MSI afterburner is not overclocked but the two identical cards vary in gpu clock and memory frequencies at standard. silicon lottery applies as much to gpus as it does to cpus. Nearly 900 units of electric this month and i am only living in one room. A bit excessive i know but it is getting colder. CPU temps on full load all 24 cores is about 65 degrees the gpus @ 100% load sit at about 55 degrees so the watercooling system is doing its job a bit too well so i don't think it has anything to do with that. maybe it is as you say just a bad batch of units.
Image

Post Reply