2020 Formula Boinc Sprints

Page 17 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Orange Kid

Elite Member
Oct 9, 1999
4,323
2,110
146

TennesseeTony

Elite Member
Aug 2, 2003
4,204
3,631
136
www.google.com
I am about to run out of tasks on one of my Linux 3950X machines, I get "[error] No close tag from the scheduler"

I switched from wifi to wired, and rebooted. No change. Ideas? 835+ tasks waiting to be reported. :(
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
@TennesseeTony, as you may have already gathered from @Fardringle's links, these results were already reported — successfully from the server's point of view (which is what matters) but unsuccessfully from the client's point of view (which is a nuisance, and is only cleared by a project reset).

BTW, I have been keeping max_tasks_reported>200</max_tasks_reported> in all of my cc_config.xml's for many months now. Most projects don't need it, but it generally does not hurt, and it prevents trouble at projects with weak database performance (such as yoyo@home). Note, it's only a preventative measure, it cannot clean up the problem in which your host is now.
 
  • Like
Reactions: TennesseeTony

VietOZ

Member
Aug 3, 2019
98
139
66
I lost 12k of tasks because of that stupid error. Like @StefanR5R , I kept my max_tasks_reported at 200 but it doesn't matter with yoyo. One of my rig was set at 25 for max_tasks_reported and still got the error. Once that error triggered, pretty much all you can do is detached.
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
@VietOZ, then I am glad that I wasn't hit by this condition myself yet, this time around. (I had the same in one or more previous contests at yoyo@home.)

But chances are good that your results which the client repeatedly tries to report were not actually lost. The client is left to believe that it never succeeded to report them, but the server is repeatedly trying to tell the client that it already acknowledged these results.

This would be easier to verify if the yoyo@home web server would show proper results tables which can be filtered like everywhere else. :-|
 
  • Like
Reactions: Fardringle

VietOZ

Member
Aug 3, 2019
98
139
66
a couple of guys on XS found that their tasks did indeed got accepted by the server. My case was out of 32k of tasks, only 20k was credited ... 12k gone.
 

TennesseeTony

Elite Member
Aug 2, 2003
4,204
3,631
136
www.google.com
Good seeing you Oz....sorry for the loss.

***************************

Thanks all for the replies. I have no idea if the work was lost or not, too lazy to sift through the odd and non standard results page offered by yoyo. I gave up after switching to the PG challenge for a bit (because, hey, [non-Canadian] bacon, how could I resist?), and reset the project.

**************************

One needs to keep an eye on sub-project ecm.....run-times seem to be getting longer. 'Normal' is 1 to 3 hours, depending on processor, but I have been getting a lot of 9 to 28 hour tasks....
 

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,219
3,798
75
So, I switched my computer over from PrimeGrid. Got 4 ECM WUs. That used 6GB each, but that's fine on this computer. Then...nothing. The server seems to be out of ECM. Another computer with less RAM is running siever for now.
 

Orange Kid

Elite Member
Oct 9, 1999
4,323
2,110
146
Yoyo@home day two

1XtremeSystems2515,438,290
2TeAm AnandTech184,817,258
3UK BOINC Team153,920,155
4The Scottish Boinc Team123,321,416
5Gridcoin103,087,187
6Rechenkraft.net81,706,529
7Planet 3DNow!6686,035
8L'Alliance Francophone4559,636
9Dutch Power Cows2512,813
10LITOMYSL Boinc Team Czech Republic1371,332

From FreeDC

xii5ku1,953,076
crashtech776,607
biodoc539,055
10esseeTony529,031
Icecold-Team Anandtech384,588
Fardringle237,939
emoga193,251
[H]Skillz95,076
[TA]OrangeKid50,191
waffleironhead18,383
Kiska5,311
geecee204
 

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,219
3,798
75
Finally got some ECM. Which is good, because Siever has quorum 2. :(
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
One needs to keep an eye on sub-project ecm.....run-times seem to be getting longer. 'Normal' is 1 to 3 hours, depending on processor, but I have been getting a lot of 9 to 28 hour tasks....
There are different batches. It is possible that a batch with long tasks is succeeded by one with small tasks. IOW run times are sometimes increasing, but other times decreasing — and differing very widely between batches.

So, I switched my computer over from PrimeGrid. Got 4 ECM WUs. That used 6GB each, but that's fine on this computer. Then...nothing. The server seems to be out of ECM.
Finally got some ECM.
I think they are generated all the time, although it may take the server a good while to catch up with demand.

I lost 12k of tasks because of that stupid error. Like @StefanR5R , I kept my max_tasks_reported at 200 but it doesn't matter with yoyo.
In retrospect, I suppose the first four or five hours of this sprint were not a good time to report results. The server might not have been quite responsive even with client-side max_tasks_reported limit, given all the scheduler requests which happened at the time, and the database size still being inflated from everything which was bunkered during the days before the sprint. (Which may not have been extraordinarily much in the grander picture, but probably very much relative to the limited capabilities of the yoyo@home server.)

@VietOZ, then I am glad that I wasn't hit by this condition myself yet, this time around. (I had the same in one or more previous contests at yoyo@home.)
I reported first results many hours after the start, at a time when everything had calmed down again. (Not due to my foresight, but just because the initial workqueues of my computers happened to last until then.) Perhaps this staying out of the "rush hour" helped to avoid this failure mode.
 
Last edited:

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
The third 19 hours:

3rd 19 hours.png

Between Saturday, 23:00 UTC and Sunday, 05:00 UTC, the yoyo@home server built up a transitioner backlog of a few hours for an unknown reason. This cleared until about 06:00 UTC.

And no, there won't be a fourth 19 hours. We have got 14⅔ hours remaining.
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
One of my hosts got into the same "[error] No close tag in scheduler reply" trouble now too. :-(
I'll watch for a good moment when to abort the remaining tasks, report all tasks (with max_tasks_reported increased temporarily to more than the number of tasks to report), and then reset the project.
 

VietOZ

Member
Aug 3, 2019
98
139
66
I reported first results many hours after the start, at a time when everything had calmed down again. (Not due to my foresight, but just because the initial workqueues of my computers happened to last until then.) Perhaps this staying out of the "rush hour" helped to avoid this failure mode.
Tony was releasing during "normal hour" and still got hit with tag error. Yoyo's admin said it's Boinc's fault, but I don't really buy it neither because it seems only happened with yoyo only. At least I've never experienced with any other project I ran. There's gotta be something else
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
Yoyo's admin said it's Boinc's fault, but I don't really buy it neither because it seems only happened with yoyo only. At least I've never experienced with any other project I ran. There's gotta be something else
AFAICT, there is a bug in BOINC *and* a problem with yoyo@home.

The bug in BOINC is that the server may generate scheduler replies which the client may not be able to pares. (Or in other words, that the client's parser is unable to parse certain replies of the server.)

The problem with yoyo@home is that the server performance is — how to say it nicely — limited (from what I last heard about it). For instance, a client may send a request in which it reports some results. The server may start to work on this request and take the results over into the database (for the validator to work on later), but it may so busy at that time, that it takes very long to get the response back to the client. So long in fact, that the client already hit the transaction timeout for this request and considers the request failed.

And therefore, the client will report the same results again later. But when it does it, the yoyo@home server sends the kind of reply which is malformed according to how the client's parser is coded.

Now there is a possibility: *Maybe* the BOINC server code was already fixed to generate only simple replies which the client code is able to parse. *If* the server was fixed in this way, then this fix did not make it yet onto yoyo@home's server installation.

[I should post my thoughts at the rechenkraft.net forum rather than here, but am still too lazy to create an account there.]
 
  • Like
Reactions: biodoc and VietOZ

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
While we've still got some more sprinting to do, I got impatient and already looked up what we have found so far:
  • @crashtech found a 49 digits long ECM factor in the OddPerfect subproject.
  • My other self found a 61 digits long ECM factor in the CullenWoodall subproject.
 
  • Like
Reactions: Ken g6 and biodoc

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
XS speculated successfully on the sprint project. The graph shows the first two of four hours of their initial dump. Since then they are running normally with 150...300 k/h, which is quite a bit better than every other team. That is, their speculative bunkering is not instrumental to their probable victory, but it gives them the opportunity to quit this sprint early and turn to whatever interests them more.
XS missed the opportunity. Instead they reverted into 2019 season mode and switched to TSBT.


Meanwhile, Formula BOINC's site code can't access its database since about UTC 00:00:
formula-boinc.org said:
MOD_PYTHON ERROR

ProcessId: 668
Interpreter: 'formula-boinc.org'

ServerName: 'formula-boinc.org'
DocumentRoot: '/web/sebastien/Formula_Boinc'

URI: '/index.py'
Location: None
Directory: '/web/sebastien/Formula_Boinc/'
Filename: '/web/sebastien/Formula_Boinc/index.py'
PathInfo: None

Phase: 'PythonHandler'
Handler: 'mod_python.publisher'

Traceback (most recent call last):

File "/usr/lib/python2.7/dist-packages/mod_python/importer.py", line 1537, in HandlerDispatch
default=default_handler, arg=req, silent=hlist.silent)

File "/usr/lib/python2.7/dist-packages/mod_python/importer.py", line 1229, in _process_target
result = _execute_target(config, req, object, arg)

File "/usr/lib/python2.7/dist-packages/mod_python/importer.py", line 1128, in _execute_target
result = object(arg)

File "/usr/lib/python2.7/dist-packages/mod_python/publisher.py", line 213, in handler
published = publish_object(req, object)

File "/usr/lib/python2.7/dist-packages/mod_python/publisher.py", line 425, in publish_object
return publish_object(req,util.apply_fs_data(object, req.form, req=req))

File "/usr/lib/python2.7/dist-packages/mod_python/util.py", line 554, in apply_fs_data
return object(**args)

File "/web/sebastien/Formula_Boinc/index.py", line 38, in index
con=MySQLdb.connect(host="localhost",user="consultation", passwd="
edited", db="stats_equipes")

File "/usr/lib/python2.7/dist-packages/MySQLdb/__init__.py", line 81, in Connect
return Connection(*args, **kwargs)

File "/usr/lib/python2.7/dist-packages/MySQLdb/connections.py", line 187, in __init__
super(Connection, self).__init__(*args, **kwargs2)

OperationalError: (1040, 'Too many connections')


MODULE CACHE DETAILS

Accessed: Mon Sep 7 02:21:57 2020
Generation: 2

_mp_ba107259ee2bb5edd44e0c0282d88147 {
FileName: '/web/sebastien/Formula_Boinc/index.py'
Instance: 1
Generation: 1
Modified: Wed Apr 1 18:34:23 2020
Imported: Mon Sep 7 02:07:36 2020
}


Hence, the last-15-hours graph is cut short:
last 15 hours.png
 
Last edited:

Orange Kid

Elite Member
Oct 9, 1999
4,323
2,110
146
Yoyo@Home final

1XtremeSystems2520,957,793
2The Scottish Boinc Team189,647,225
3TeAm AnandTech158,331,229
4UK BOINC Team127,537,745
5Gridcoin104,571,074
6Rechenkraft.net82,975,170
7Dutch Power Cows61,403,060
8L'Alliance Francophone4970,886
9Planet 3DNow!2953,922
10LITOMYSL Boinc Team Czech Republic1750,116

From FreeDC

xii5ku3,759,166
crashtech1,102,031
10esseeTony957,637
biodoc796,630
Icecold-Team Anandtech613,409
Fardringle308,127
emoga294,231
[H]Skillz137,837
[TA]OrangeKid123,663
Ken_g6
35,901​
waffleironhead25,298
Kiska23,830
geecee1,483

Results from Italy o_O

  • Pierre GASLY1:47:06.056
  • Carlos SAINZ+0.415
  • Lance STROLL+3.358
 
Last edited:

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
Another F1 race went down in Italy last weekend:
Lewis Hamilton ....... 2:19:35.060
Valtteri Bottas .................. +4.88
Alexander Albon .............. +8.064​

Next weekend, there will be an FB sprint but no F1 race.
After that, the F1 and FB schedules are aligned again.