Posts by chip

21) Message boards : Number crunching : Minimum quorum 3, initial replication 2 WTF? (Message 249)
Posted 10 Jul 2020 by Profile chip
Post:
Is there any chance to be validated for app. 2.03 tasks, or is better to set NNT?


Solved this, it was an oversight. Thanks
22) Message boards : Number crunching : Version 2.03 Estimated time to complete (Message 248)
Posted 10 Jul 2020 by Profile chip
Post:
Yeah, the estimates we plug in for the initial estimate are way off.
The app itself should feed an accurate estimate back to the client, but it'll take some time to tune itself I imagine.

We'll work on providing a more accurate initial estimate soon.
23) Message boards : Science : Cactus Run & Beyond? (Message 247)
Posted 10 Jul 2020 by Profile chip
Post:
Throwing into Google:
- https://packpng.com/
- https://www.youtube.com/watch?v=eC7f9tMslVE


So, what's the transition plan? How long are we going to hunt for cacti and where to next?


Most of the activity is happening on our Discord, some pending projects are finding the seed which was used for the main menu's panorama image, and finding large emerald ore veins; pack.png is also pending but we're relying on an external team to provide us with the code. We've discussed wanting to incrementally build capability to take any screenshot and locating the seed which can be used to generate that same world, small parts of each project feed into each other.
I expect we'll have more and more common libraries and functions which will allow us to rapidly spin up new projects based on the community's feedback and ideas.
24) Message boards : Number crunching : These are not badges! (Message 246)
Posted 10 Jul 2020 by Profile chip
Post:
Yes, the only permanent badge so far is the 'Founder' badge for taking part in the beta-testing.
I don't believe there's a difference in nomenclature for the the transient/ephemeral badges you describe.

I will be introducing more 'permanent' badges in the coming days.
25) Message boards : News : Kaktwoos 2.03 and Badges! (Message 234)
Posted 8 Jul 2020 by Profile chip
Post:
Now please an option in community preferences for showing/not showing of badges. As I don't need to see them on every post and even in the forum index as well. Clutters. Of course I can go block them in uBlock Origin, but it would be friendlier to have the option to not show them. And leave that up to the user to decide.


Is there some code I can steal for this, or am I missing a config toggle?
26) Message boards : News : Kaktwoos 2.03 and Badges! (Message 233)
Posted 8 Jul 2020 by Profile chip
Post:
Also, we now have badges!

it would be better if they were not :(

"Top %%" is initially in the project, no need to change the standard (familiar) image
They are dynamic (depending on the speed and not on the work performed) and makes sense only within the project.
They are not saved by statistics sites and are not used in signatures.

The badge "Finder" is acceptable, it’s only advisable to pick up a picture that is associated with the project - if you look at them against the background of huge collections of badge collectors ;)

If it’s still a badge "project founder" then it is better to use special roles on the forum


Ah, there were no images bundled in the containerized version of BOINC. We're rebuilding ourselves for 1.2.1, so I'll use the defaults at that point assuming they're part of the usual BOINC installation.

The 'Founder' badge is to indicate users who joined during the beta-testing. I could rename it to something more obvious.
27) Message boards : News : Kaktwoos 2.03 and Badges! (Message 230)
Posted 8 Jul 2020 by Profile chip
Post:
You may have seen many workunits just disappear from existence today.



Worry not! We realised there were far too many workunits scanning duplicate seeds, so we've scaled back the original workunits to the correct seed ranges (no in-progress results were touched, so none of you should have lost any credit).

We have located some promising seed candidates which were missed from processing and can be used as an input to this job, so they are currently set as the highest priority.

Also, we now have badges! If you view the forums, any comments in threads, or on the leaderboards; you will see the new badges.
We're open to suggestions for future badges, so please leave us some comments on this thread.
28) Message boards : Number crunching : Computation errors on CentOS 7.x (Message 228)
Posted 8 Jul 2020 by Profile chip
Post:
Fixed in latest update
29) Message boards : Number crunching : Badges. And Stats Export (Message 227)
Posted 8 Jul 2020 by Profile chip
Post:
Badges now exist, we'll be expanding on them with time, but I hope these happy badges will make happy people :)
30) Questions and Answers : Web site : House keeping (Message 226)
Posted 8 Jul 2020 by Profile chip
Post:
By the way chip, you can also use the BOINCNetwork on Discord: https://discord.gg/wPRafUq


Funnily enough I'm already reasonably active in that Discord, they've been helpful for a few queries
31) Questions and Answers : Web site : House keeping (Message 221)
Posted 7 Jul 2020 by Profile chip
Post:
it is necessary to regenerate dead WU - they are normal data and might contain solutions


We'll definitely reinstate any failed workunits. I first need to determine if it's safe to remove the error mask and change the thresholds on the failed workunits rather than re-creating them.

It's causing my computer to not be eligible to get work. If that's what you want, then so be it.

at the wrong time, it’s too early to ban hosts

You may run the project as you seem to find fit, I don't mind. But just filling in values without reading the documentation isn't going to be of help here. And it's not needed. I asked you a couple of times already to email the BOINC Projects email list if you feel you're in over your head, as the other project admins are more than willing to help out.


That's a surprise and wasn't my intention. I don't want to ban hosts. The purpose of the change was to increase the priority of duplicate jobs so that they'd be processed by reliable hosts; this was to ensure incomplete workunits are prioritized to complete validation as per this thread.

I thought I fully understood the documentation, my comprehension of the options are that reliable hosts are selected first to run work which have a priority equal to or higher than the value provided by <reliable_on_priority>.
Results with priority at least reliable_on_priority are treated as "need-reliable". They'll be sent preferentially to reliable hosts.


Do know that anyone with a new computer signing up now can't get work as their computer isn't reliable. And it's not explained either in the scheduler messages. Mine just says it asked for work but didn't get any.

This may have been due to a failed version bump of the kaktwoos app to 2.02 which has since been reverted. It looks as though all computers are recieving work now as far as I can see from the data. I may be mistaken though, in which case we can look at reverting another change to resolve this.

Trouble is, my host hasn't had the chance yet to reach 10 consecutive valid results to be allowed to be reliable. 9 is the most it had in a row before there were two that couldn't be validated due to the problem I pointed out.

first get a stable application and distribution to suitable hosts.

can set the task the following parameters


error = total, can be reduced
success = quorum + 1 (accept too late, but successful)


Kaktwoos 201 has 9273 successes and 2550 failures, 2368 of the 2550 failures are attributable to 6 problem hosts, most of which do not have OpenCL headers installed.

hostid  num_fails  reason
302     481    nvidia opencl headers missing
471     424    HD 4000, now specifically disabled due to lack of OpenCL 2.0 compat.
1196    413   old APU, incompatible drivers
510     405    old drivers, user couldn't upgrade, now disabled
1436    375   outdated drivers
1035    272   missing libraries which we need to statically compile see (https://minecraftathome.com/minecrafthome/forum_thread.php?id=25)


We're working on a solution to send a message to the user via the scheduler in situations where hosts appear to have the correct drivers and match the plan class, but do not have the correct drivers.

I will increase the total error count on the workunits to account for this instability though which should be a band aid in the meantime.


        
32) Message boards : Number crunching : Computation errors on CentOS 7.x (Message 220)
Posted 7 Jul 2020 by Profile chip
Post:
I have checked binaries which you sent to me, and they still loads this lib dynamically. Please link app with -static-libstdc++ option, it should help.


Hi Daniel, bare with us I'll see about pushing the new ones to you for testing before rolling them globally.
33) Questions and Answers : Web site : House keeping (Message 214)
Posted 6 Jul 2020 by Profile chip
Post:
Two tasks that didn't validate due to too many errors:
https://minecraftathome.com/minecrafthome/workunit.php?wuid=1257094
https://minecraftathome.com/minecrafthome/workunit.php?wuid=1257331

Now, max # of error/total/success tasks = 3, 10, 6
This means a maximum of 3 errors can be had, which in both tasks is 4 (4 independent computers returned it with computing error).

331 I see I got yesterday evening, I ran it to 50% then shut down. Continued it today, in the mean time a couple of Intel GPUs erred on it. I was the 6th to return it but it already had 3 (and even 4) errors.
094 is a different story though, i t had erred 3 times already when it was sent out to me, I returned it correctly, one other computer erred on it as well. Why wasn't it sent out another time, or why was it sent to me in the first place if max error # is 3?


Enabled the reliable hosts stuff, should solve it for now
34) Message boards : Number crunching : Computation errors on CentOS 7.x (Message 213)
Posted 6 Jul 2020 by Profile chip
Post:
CentOS 7.x (and RedHat 7.x too) uses older package versions. New version of cactus app which you added few days ago no longer is linked statically with system libraries, and it fails on my machine in following way:
../../projects/minecraftathome.com_minecrafthome/kaktwoos_2.01_opencl_nvidia: /lib64/libstdc++.so.6: version `CXXABI_1.3.9' not found (required by ../../projects/minecraftathome.com_minecrafthome/kaktwoos_2.01_opencl_nvidia)
../../projects/minecraftathome.com_minecrafthome/kaktwoos_2.01_opencl_nvidia: /lib64/libstdc++.so.6: version `GLIBCXX_3.4.21' not found (required by ../../projects/minecraftathome.com_minecrafthome/kaktwoos_2.01_opencl_nvidia)

Example failed WU https://minecraftathome.com/minecrafthome/result.php?resultid=2715670

Please recompile your app with libraries linked statically.


The issue was slightly different than I expected here, let us know if the latest app kaktwoos 2.02 works for you. If not, I'll try to rapidly push a new fix.
Thanks
35) Message boards : Number crunching : Dual IDs (Message 212)
Posted 6 Jul 2020 by Profile chip
Post:
The problem is that I don't know the email address of the founder account not how and when it got created.


I private messaged you the founder email
36) Message boards : Number crunching : Getting 0 tasks on tahiti gpus. (Message 211)
Posted 6 Jul 2020 by Profile chip
Post:
Does this project require a specific OpenCL version? I believe these cards support 1.2.

Thanks.


You can find our plan classes here.
37) Message boards : Number crunching : deadline is too short (Message 210)
Posted 6 Jul 2020 by Profile chip
Post:
Yeah, the iGPU work disabling is on BOINC's side. The hardware can be kept active and used for other projects if you want


You can exclude a specific GPUs in your cc_config like this:
<exclude_gpu>
   <url>https://minecraftathome.com/minecrafthome/</url>
   <type>intel_gpu</type>
</exclude_gpu>


However, I have increased the deadline on all jobs by 3x to 72 hours and reliable hosts which normally turn work around in under 4 hours will receive priority work with a 36-hour deadline (duplicate results, usually retries or errors from incomplete workunits).
38) Message boards : Number crunching : deadline is too short (Message 201)
Posted 5 Jul 2020 by Profile chip
Post:
Yeah, I think we're going to experiment wiht multi-size, those Intel GPU jobs take their sweet time.
Either that, or I'll just simply bump the deadlines up to 72 hours or something
39) Message boards : Science : Cactus Run & Beyond? (Message 185)
Posted 4 Jul 2020 by Profile chip
Post:
Pack.PNG is the ultimate goal
40) Message boards : Number crunching : Dual IDs (Message 184)
Posted 4 Jul 2020 by Profile chip
Post:
Happy to assume good faith here if all else fails


Previous 20 · Next 20