New Beta Test starting Oct 31, 2013

[D_A]

We will be conducting a new beta test starting today. This test will have system requirements of:

400MB Ram
512MB of disk space
10MB result files

The initial beta test will have 5000 work units, with a quorum of 2. This means that 10,000 results will be sent out. I will be sending them out slowly so they could take a few hours to distribute. This test will be for Windows, Mac and Linux. Android will not be included in this beta test.

Thanks,
-Uplinger

No word yet on 32 or 64bit, though I doubt 64bit linux will be in this first beta test.

Edit: 32 and 64bit on Windows and Linux, 64bit only on Mac.

[Evantaur]

got a few :up:

[anubis]

None for me :(

[D_A]

I just got the one. Might be a longer second stage beta, or it might go into production soon. We'll find out in due time I suppose.

[Evantaur]

World Community Grid 7.19 Beta Test BETA_BETA_9999985_0263_0 00:00:59 (00:00:57) 98,05 0,500 07:21:24 10.11.2013 8:28:36 Running hellhound
World Community Grid 7.19 Beta Test BETA_BETA_9999988_0117_0 00:00:59 (00:00:57) 97,55 0,500 07:21:24 10.11.2013 9:28:30 Running hellhound


7 hours 21 minutes :shocked:

[PoppaGeek]

got 2 64 bit linux, est. run time 10:31 Opteron 8356 @ 2.3

Kinda big for HCC? :shrug:

[PoppaGeek]

For those not following the thread at WCG the BETAS have a "quirk"


First Beta WU finished after 1 hour 55 minutes.

The WU having 0.500% progress after 45 minutes and then jumped to 20.000% 10 minutes later, jumped to 40.000%. So it seems to be no bug - just let those WUs crunch undisturbed.
Edit: Another jump to 60.000% just happened.

Do not worry if they seem stuck. :up:

EDIT:
I have one "stuck" and 2 progressing normally(I got another one).

[PoppaGeek]

Link (https://secure.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=436340)


First thanks to all of the Beta testers. This new application has a larger number of different types of workunits than normal. This results in several possible paths through the code and some variance in workunit completion time and time between checkpoints.

The issue with workunits being stuck at 0.5% is an issue with the percent complete calculation. If the application is using cpu time it is probably not stuck so please let it continue to run for a while to verify. The 0.5% is some initialization which is taking longer for some of the types of workunits. We will correct this and determine a better way to show progress during the initialization.

The application should checkpoint at most every ten minutes. This is because some of the types of workunits produce larger checkpoint files and this will help limit disk I/O. Every workunit should checkpoint so we are looking into the ones that are not checkpointing.

The large output file errors are an issue with this being new research. We anticipated a smaller percentage of these so I apologize there are more than anticipated. The applicaitoin includes input parameters to limit the output sent back to the most critical and these parameters will be enhanced as we go. The next beta should have a very minimal amount of these. All users are getting credit for these errors.

Thanks again for all the help in getting new research running.

Thanks,
armstrdj

[Evantaur]

Task Host CPUtime Rtime points PPH | PPD
-----------------------------------------------------------------------------------------------
BETA hellhound 1.71 1.71 38.3 22.39 | 537.36
Average runtime per task: 1.71 h Average Points per hour: 22.39


looks like not worth it :P

[PoppaGeek]

2 finished a little over 5 hours each.

third went 6:30 hours

[Rob_B]

I have a few validated that were done in 2-3 hours and one in just 10 minutes. But also had 3 errors already.
Funny thing is that I'm my own wingman on one of the wu's that errored :)

[Evantaur]

got 4 of them stuck :P

[D_A]

Mine died. :( Oh well ... problems are being found which means they get fixed and we get more beta units! :)

[[XC] Oj101]

Not one here :(

[OldChap]

16 + a re-send here but on the smaller rigs nothing for the 2 32's

Huge disparity in times projected and in running for some wu's

The re-send errored 131 for the primary user so I expect the same on that one but I only brought mine to the top a few hours ago and am waiting on the first completion.

Had one wu that sat at 0.5% for 2h30m then it jumped to 50% (where it has stayed since)

Getting 99% plus efficiency with these

http://www.lakecityquietpills.com/photo/multihost/images/92655495971280765031.png (http://www.lakecityquietpills.com/photo/multihost/)

an hour later:

http://www.lakecityquietpills.com/photo/multihost/images/88927144893783083966.png (http://www.lakecityquietpills.com/photo/multihost/)

[jeanguy2]

Only got one FAAH beta 7.19 (beta_beta_9999987_0336_0) WU on a 2600K WIN 7 cruncher. It ran for 03:56:30 but then produced an upload error on completion. :(

Details below:

Run complete, CPU time: 13919.860029
11:48:26 (23596): called boinc_finish

</stderr_txt>
<message>
upload failure: <file_xfer_error>
<file_name>BETA_BETA_9999987_0336_0_0</file_name>
<error_code>-131</error_code>
</file_xfer_error>

[0ne.shot]

I haven't seen a beta WU on my rigs yet.

[OldChap]

Only 10,000 in the first tranch and quite a few problems to fix so I am sure that there will be quite a few more

Patience you must have http://www.pic4ever.com/images/129fs238648.gif

(and betas turned on in preferences)

[NKrader]

still have none... :( :( :(

[yojimbo197]

still have none... :( :( :(

Me neither. I'm beginning to wonder if its discrimination against Linux.....
:mad:

[OldChap]

Batch 1 were issued in the first hour and a half with re-sends spread over about 24 hours..... Be lucky for batch 2 when they test their fixes

[D_A]

Me neither. I'm beginning to wonder if its discrimination against Linux.....
:mad:

Nope, just a very limited number of units all round. 10,000 total from memory. I got one. I would have liked a queue full, but that just ain't gonna happen.

[Evantaur]

looks like they ran out of FAHH tasks too :shrug:

[OldChap]

I can't tell because I am working through a whole batch of work that is more difficult, runs longer, gives less points and is generally a PITA. When this happens then the rigs go into High priority mode and don't fetch new work (not needed due to longer run times). WCG work schedulers are up though.

[nanoprobe]

The 10,000 original tasks were gone jn 1.5 hours. The beta started around midnight EST so many of us missed them. I got 21 resends, they all failed but I still got credit so not all was in vain.

[OldChap]

NEW BATCH This time with 2 day reporting time

earliest start time for me was 2h33m ago

[[XC] Oj101]

I suppose my eight day queue has prevented me from getting any work units :(

[PoppaGeek]

Got 7, 5 Windows 2 Linux

[OldChap]

Oj101;5214858']I suppose my eight day queue has prevented me from getting any work units :(

I have 3 rigs running in high priority mode that received some... I think your rig just needs to be asking for work at the right time

I am still getting primary work in dribs and drabs ....have already had re-sends too

EDIT: Here's the reason my 4th rig got none so far:

spiders

119736 World Community Grid 04/11/2013 22:29:30 Sending scheduler request: To report completed tasks.
119737 World Community Grid 04/11/2013 22:29:30 Reporting 2 completed tasks
119738 World Community Grid 04/11/2013 22:29:30 Not requesting tasks: don't need
119739 World Community Grid 04/11/2013 22:29:33 Scheduler request completed

[Rob_B]

I got 14 beta wu's spread over 3 rigs. And all run in high priority.

But I'm running all rigs with a small cache. Just 1 day + 0.3 additional.

[PoppaGeek]

Mine are all running High Priority and the cache on all mine are less than 2 days.

[PoppaGeek]

Got 4 more Windows. So far 4 valid, 3 errors.

All but 3 of the betas I have gotten so far show no errors by me and wingman. All other have at least 1 error some several.

[OldChap]

30 wu's
shortest 0.29hrs runtime
longest 12.59hrs runtime
average 3.77hrs runtime
valid 23
pv 7
errors 0 (so far)
max pphour (3700k) 29.7
min pphour (e5-26xx) 9.95

111 hours of runtime (4.5 days+) towards my gold beta badge :D

[OldChap]

Got another new batch on the spiders rig this time.... it got missed yesterday

This is beta 7.22. you may also see re-sends from 7.21

Also, the new work units will end with "b_#" where # is some integer between 0 and 9.

[[XC] Oj101]

Can you tell me the "Application" or "Name" as reported by BOINC? I have a MASSIVE queue (I messed up the settings :p:) and I want to arrange either column by name and see if I was lucky enough to be sent some.

[D_A]

The Application is "Beta Test", and the unit names start with "BETA_BETA"

[D_A]

Had to extend my cache out a few days but I got four! W00T!

[OldChap]

I pushed mine to 4 days from 3 but the hard wu's I had over the last few days have put mine into high p already, still.... got 23 and 11 on the big rigs so far and still getting more for now. Sure helps to have wu's finishing close together and I aborted a couple of CEP's... just to make space of course ;)

[anubis]

Got two

Name BETA_BETA_9999983_0519b_0
Application Beta Test 7.22
Workunit name BETA_BETA_9999983_0519b
State Running
Received 6.11.2013 0:29:33
Report deadline 8.11.2013 0:29:32
Estimated app speed 1,36 GFLOPs/sec
Estimated task size 71397 GFLOPs
CPU time at last checkpoint 00:00:00
CPU time 00:05:11
Elapsed time 00:05:14
Estimated time remaining 14:27:34

Name BETA_BETA_9999984_0318b_0
Application Beta Test 7.22
Workunit name BETA_BETA_9999984_0318b
State Running
Received 6.11.2013 1:00:39
Report deadline 8.11.2013 1:00:38
Estimated app speed 1,36 GFLOPs/sec
Estimated task size 71397 GFLOPs
CPU time at last checkpoint 00:00:00
CPU time 00:06:26
Elapsed time 00:06:31
Estimated time remaining 14:26:18

[OldChap]

A little something for you guys just before I head for bed...

You know how Dave says a picture paints a thousand words....I Managed a full loadout across both Spiders and Arachnid:

http://www.lakecityquietpills.com/photo/multihost/images/88484537751006208407.png (http://www.lakecityquietpills.com/photo/multihost/)

[Evantaur]

how the f? :O

[D_A]

The scheduler won't give you any more than it's certain you can complete by the VERY short deadline.

[Evantaur]

http://nikita.tnnet.fi/~evntr/boinctasksd.jpg

wee!! :p:

[[XC] Oj101]

And I STILL haven't received ONE! :(

[0ne.shot]

I just got 2 Beta Test 7.22 WUs!!!! It looks like someone is hogging them all!

[BlindFreddie]

All the v7.22 betas I got on Win7-x64 errored out. They start, run for about 10 sec, do an "exit(0)", restart, and repeat this until they hit a limit on the no of "exit(0)s" allowed (100).
My 1 machine that does not run w7-x64 runs XP-x64 and it ran the 1 short WU it got OK.
Think it's a bug in Win7-x64, surprise, surprise.

Don't know whether 32-bit XP vs W7 is similar.

Linux is reported as OK, Mac probably OK too.
----
Guys, before you go loading up on betas, please remember that WCG need and want diversity in hardware and OSes when they run beta tests.
In this case there are a number of different WU types too, so it won't hurt if you run a reasonable number of beta WUs under each of your hw/OS combos.
But doing things that detract from the results of the beta tests for the sake of a different-coloured little square on your WCG forum posts hardly fits in with why we are crunching for WCG.

But I do like the colour of my gold β badge :D

[D_A]

They have fixed numbers of units for each OS and variation because they're all coded slightly differently. Windows units don't run on Linux and v/v. WCG isn't like Test4Theory where the work units run in consistent virtual machines and the host OS isn't relevant, ours run on bare metal. It's not possible for one person with a huge farm of 64bit Linux machines to vacuum up all the units that could have gone to Windows or Mac.

[PoppaGeek]

Seems they must have the Linux version sorted out as I see few errors. Windows is still buggered from the looks of it.

[OldChap]

7.24 now as they are working through the issues


New Beta Test starting Nov 6, 2013 v7.24

This beta test is testing if we have fixed the issue with the too many exits issue.

This issue appears to have been caused by a change in the BOINC API layer for the research application. Every 10 seconds, the research application will check to see if the BOINC core client is still running. Previously, this check was done by iterating through all the running processes on the system to see if the process id could be found. This was changed to use the OpenProcess windows api call. The BOINC API treated any failure of this call as the core client not being there. However, beta test 7.23 confirmed that in the cases we were seeing it was returning "OpenProcess failed with error 5: Access is denied.". This represents a security policy issue rather than the core client not running. We suspect this has to do with the permission and user structure used when running BOINC as a service. We will pass this information along to BOINC for them to review and decide what to do.

In the meantime, we reverted this portion of the API back to the previous mechanism and have released this as version 7.24. We are hoping for a cleaner run this time.

[jeanguy2]

Got 2 new Beta WUs (beta17_7.22) on a 3930k Win7 rig Both processed OK


Result Name Device Name Status Sent Time Time Due / Return Time CPU Time / Elapsed Time (hours) Claimed/ Granted BOINC Credit
BETA_ BETA_ 9999986_ 0819b_ 0-- ginette Valid 11/5/13 23:40:29 11/6/13 09:59:08 2.38 / 2.38 85.2 / 84.3
BETA_ BETA_ 9999986_ 0803b_ 0-- ginette Valid 11/5/13 23:40:29 11/6/13 09:59:08 0.48 / 0.48 17.3 / 19.5

[0ne.shot]

I just woke up and updated the project and received two 7.21 Beta WUs.

[Rob_B]

I have one 7.21 wu and 25 7.24 wu's on 4 rigs.

The last batch I got on the 4th all validated without errors.

[PoppaGeek]

We are running another expanded round of testing using the same application version. There are roughly 2.5 times as many tasks in this round as the previous. We have also given the beta more weight so it should be easier to get more tasks.

It is available on 32 & 64 bit Linux, 64 bit Intel Mac, 32 & 64 bit Windows.

knreed -Nov 7, 2013 6:54:49 AM- Re: New Beta Test starting Nov 6, 2013 v7.24 (http://www.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=437312)


EDIT:

Have 10 already, 5 Win and 6 linux. :up:

[jeanguy2]

I got the Beta 7.24 WUs on all four of my rigs for a total of 32 WUs.

The number of Beta 7.24 downloaded by each rigs matches the # of thread available on their individual CPU.

[PoppaGeek]

I am up to 17 now, several finished today, more keep coming. Not seeing any errors so far. :up:

[Rob_B]

Only 14 left at the moment. About an hour ago I had all my 32 cores crunching beta wu's. And no errors here either.

If they keep going like this I will get my silver badge after all. Still 15 days to go.

[PoppaGeek]

Interesting memory footprint on these run from 43 megs to 122 megs and everything in between. Windows usually the highest. :shrug:

[anubis]

Got a batch of 12WUs 7.24's, so yes, matches the core count of cruncher. No errors :up:

[OldChap]

Need about another 2 days and 2 hours of these for a gold badge but the last issue seems to run ok for me.... hope that is not the end of testing

[PoppaGeek]

Still have 5 pending validation but all others valid. Others in forums reporting high success rate. Should be getting close to release. :)

[OldChap]

Well, based on the wu's I got for this beta I reckon I need about another 16 new projects before I get my sapphire :D

I wonder if MCM will lend itself to having a Vina version

[Rob_B]

For a moment I thought we got a new beta project. But I just got a handful of 7.24 resends.

[OldChap]

got about 18 hours worth here ... need a few hours more for gold

[Rob_B]

I could use those. I need 6 days more for silver.