Pages: [1]
Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-03-30 20:54:39

I had this part of https://boincstats.com/en/forum/18/11504, but seems unrelated now that the cert issue is fixed.

Yesterday I had changed the SHAREs of all my projects on BAM. They used to be all at 100, now they are at 1/10/25/50/100/200/500 levels. All project sites seem to have gotten the new numbers; good. However, running TeamMgrUpdate as well as an UPDATE on ALL projects on EVERY host, not all hosts have received these new numbers from the projects (nor from BAM), and for some of those that have, BAM still (after a good AcctMgrUpdate) shows some/most of the resource shares for those hosts at 100 even though the host is using a different number as I can see under the BOINCMgr/Projects tab on those hosts.

For example, https://boincstats.com/en/bam/host/629507 shows, among others:

SETI@Home 265:35:31 100
SETI@Home Beta 51:49:07 100
WUProp@Home 08:29:51 100

But, on the host, under the Project tab, the numbers are 100/50/500, respectively (and, these are the only projects this host is attached to). And, those are the correct values I have set in https://boincstats.com/en/bam/projects/. And, they have travelled to the projects and (mostly) made their way back into my hosts. But, for some hosts, AcctMgrUpdate doesn't seem to update the host-specific data in BAM. On some other hosts, this worked perfectly. And, on some other hosts, the new setting doesn't get set properly after an UPDATE with some of the projects. Mystified...

Tuna

PS: Also of note is the fact that https://boincstats.com/en/bam/host/629507 shows some 1/10/25/50 values, even for projects that this host is not even attached to, but by far less than what the reality is as reflected on the https://boincstats.com/en/bam/projects/ page (and the hosts that are attached to those projects and have it correct locally, regardless if their host page on BAM is correct).

Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-03-30 21:07:53

Also, on https://boincstats.com/en/bam/projects/, I am seeing these, instead of success:

Einstein@Home Incorrect response from project 50
ATLAS@Home Project down 10

However, looking at the projects, the values 50 & 10 have been properly set, and the projects are up. So, BAM! has no problem setting the value in these two projects, but somehow fails to hear the SUCCESS in return.

Tuna
Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-03-31 01:09:59

Tuna Ertemalp wrote:

...on some other hosts, the new setting doesn't get set properly after an UPDATE with some of the projects. Mystified...

For this subset of the problems, which is separate than the main BAM vs Host issue I am reporting, I have created https://boinc.berkeley.edu/dev/forum_thread.php?id=11541&postid=77058.

Tuna
Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-03-31 01:26:35
last modified: 2017-03-31 01:29:57

Tuna Ertemalp wrote:
But, for some hosts, AcctMgrUpdate doesn't seem to update the host-specific data in BAM.


Looking at acct_mgr_request.xml on this host, I see correct values which BAM doesn't seem to get:

-<project>
<url>http://setiathome.berkeley.edu/</url>
<project_name>SETI@home</project_name>
...
<hostid>7914231</hostid>
<attached_via_acct_mgr>0</attached_via_acct_mgr>
<resource_share>100.000000</resource_share>
</project>

-<project>
<url>http://setiweb.ssl.berkeley.edu/beta/</url>
<project_name>SETI@home Beta Test</project_name>
<hostid>77872</hostid>
<attached_via_acct_mgr>0</attached_via_acct_mgr>
<resource_share>50.000000</resource_share>
</project>

-<project>
<url>https://wuprop.boinc-af.org/</url>
<project_name>WUProp@Home</project_name>
<hostid>91315</hostid>
<attached_via_acct_mgr>0</attached_via_acct_mgr>
<resource_share>500.000000</resource_share>
</project>


But I also notice attached_via_acct_mgr=0, which is totally bogus. All my machines had started using BAM. But, even if this were to be the problem, why would BAM not reflect that setting just like it does all the other settings for each project regardless if AttachedViaAcctMgr is 0 or 1? And, if a project is not attached on a host, why show a default 100 instead of the actual ResShare set in MyProjects or a 0?

Tuna
Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-04-10 04:51:21
last modified: 2017-04-10 04:55:11

So, I PAINSTAKINGLY went through all of my 12 hosts with ~50 projects they are attached to, and made sure every project on each host has the correct resshare that I set under My Projects, and I also made sure that each project's own site also showed the same resshare in their ProjectPreferences under YourAccount. To do this, I had to identify all the projects that for some strange reason wouldn't get the new resshare from the project, drain it of any remaining tasks with NoNewTasks+AbortNotStartedWork+DelayedDetach, wait for detach, and then reattach it using BAM. So, now the MyProjects page, the projects themselves and my hosts are all in sync.

However, the resshares listed for the projects on individual host pages in BAM are not at all correct. Even though many AcctMgr syncs happened since then, BAM thinks mostly of the wrong resshare values for a project on a host. Below is just the list from one of my machines. The triple values in the parentheses are: "(the default value in My Projects, BOINCMgr on the host, Host Page on BAM)"; a "**" indicates a mismatch between 1st & 3rd value. Pretty much resshare data on all of my host pages on BAM are in this hosed state. They were hosed like this before I spent the time to fix the values on the hosts themselves, and they still are hosed after I fixed the hosts. Something is wrong.

Also, as an example, here is the acct_mgr_request.xml section for Asteroids@Home which is one of the many projects on this host that doesn't have the right data under BAM:

<project>
<url>http://asteroidsathome.net/boinc/</url>
<project_name>Asteroids@home</project_name>
<suspended_via_gui>0</suspended_via_gui>
<account_key>.........</account_key>
<hostid>265923</hostid>
<not_started_dur>42827.805119</not_started_dur>
<in_progress_dur>0.000000</in_progress_dur>
<attached_via_acct_mgr>1</attached_via_acct_mgr>
<dont_request_more_work>0</dont_request_more_work>
<detach_when_done>0</detach_when_done>
<ended>0</ended>
<resource_share>25.000000</resource_share>
</project>

ResourceShare=25, but in the below list, you'll see that BAM knows that the default value for the project is 25, but it thinks that on this particular host it is set to 100. Also note that it is AttachedViaAcctMgr=1 (as is the case for ALL the projects with "**" next to them below), so my previous analysis in this thread based on another host seems to have been was irrelevant.

And, here is the list of the mismatches for this host:

Acoustics@home (50,50,50)
Albert@Home (50,50,100**)
Amicable Numbers (50,50,50)
AndersonAttack (1,1,1)
Asteroids@home (25,25,100**)
ATLAS@Home (10,Not Attached,100**)
Bitcoin Utopia (1,Not Attached,1)
BURP (10,10,100**)
CAS@HOME (1,1,100**)
Citizen Science Grid (25,25,100**)
Climate Prediction (25,25,100**)
Collatz Conjecture (200,200,100**)
Cosmology@Home (10,10,100**)
DENIS@Home (10,10,100**)
DistributedDataMining (10,10,100**)
DrugDiscovery@Home (10,Not Attached,10)
Einstein@Home (50,50,100**)
Enigma@Home (10,10,100**)
FiND@Home (1,1,100**)
Gerasim@Home (10,10,100**)
GoofyxGrid@Home (25,25,100**)
GPUGRID (200,200,100**)
Leiden Classical (10,10,100**)
LHC@Home (100,100,100)
LHCathome-dev (5,Not Attached,100**)
MilkyWay@home (25,25,100**)
MindModeling@Home (10,10,100**)
Moo! Wrapper (200,200,100**)
NFS@Home (10,10,100**)
NumberFields@home (10,10,100**)
Primaboinca (10,10,100**)
PrimeGrid (200,200,100**)
Quake Catcher Network (1,Not Attached,1)
RADIOACTIVE@HOME (1,Not Attached,1)
RALPH@Home (5,5,100**)
RNA World (100,100,100)
Rosetta@Home (10,10,100**)
SAT@home (1,1,100**)
SETI@Home (100,100,100)
SETI@Home Beta (50,50,100**)
Sourcefinder (10,10,10)
SRBase (25,25,100**)
Stop@home (25,25,25)
SZTAKI Desktop Grid (5,5,100**)
The Lattice Project (1,Not Attached,100**)
theSkyNet POGS (25,25,100**)
TN-Grid (10,10,100**)
Universe@Home (10,10,100**)
VGTU project@Home (10,10,100**)
Volpex (1,1,100**)
WEP-M+2 Project (1,Not Attached,1)
World Community Grid (100,100,100)
WUProp@Home (500,500,100**)
YAFU (10,10,100**)
yoyo@home (10,10,100**)

Clearly, the 3rd values (from BAM host data) should match the 2nd (directly from the host). Additionally, if 2nd is "Not Attached", then the 3rd should preferably be either NaN or 0, or at the very least, whatever the 1st value is (the project's default resshare), and definitely not a generic 100.

Thanks
Tuna

Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-04-10 05:00:28

Additionally, this is still happening:

Tuna Ertemalp wrote:

Also, on https://boincstats.com/en/bam/projects/, I am seeing these, instead of success:

Einstein@Home Incorrect response from project 50
ATLAS@Home Project down 10

However, looking at the projects, the values 50 & 10 have been properly set, and the projects are up. So, BAM! has no problem setting the value in these two projects, but somehow fails to hear the SUCCESS in return.

Tuna


Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-04-14 21:20:10

And, a more recent discovery: http://www.primegrid.com/forum_thread.php?id=7395#106956

Seems changing project priorities in BAM might be resetting some of the other project preferences back to their defaults at the project sites while changing the project priority correctly. Hard to prove, but someone else reported seeing that recently, and I just checked and found that random projects had random project settings reset to NO from what I would have set to YES over the last few years.

Tuna
[BOINCstats] Willy
 
Forum moderator - Administrator - Developer - Tester - Translator
BAM!ID: 1
Joined: 2006-01-09
Posts: 9136
Credits: 348,241,820
World-rank: 2,972

2017-04-15 07:00:56

A bit of background information which might shine some light on the problem.


  • When the recourse share is changed on the My projects page only the projects are updated, nothing is changed in the database for the hosts.
  • On the host page you can also set the resource share. Setting it to '-1' tells the host to follow the resource share set at the project. Any other value sets the resource share for that specific host.
  • Your hosts have a mixture of '-1' and other values.


So, if host x has the resource share for SETI set to 100 and you change the resource share for SETI to 50 on the My projects page, host x will not follow and remain at 100.

Does this clear up things or is it even more confusing now?

Please do not PM, IM or email me for support (they will go unread/ignored). Use the forum for support.
Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-04-15 08:44:43

[BOINCstats Willy wrote:
When the recourse share is changed on the My projects page only the projects are updated, nothing is changed in the database for the hosts.

That is also totally what I would expect. But then, when a host does an Update with the project, it gets (or should get; in my case, some hosts got the new resshare from the projects, some hosts didn't) the new resshare from the project, and starts using it. At this point, during the next AcctMgrSync by that host, I would have expected that new resshare value to travel from the host back to BAM to show up in that host's projects table. Wrong expectation?


[BOINCstats Willy wrote:
On the host page you can also set the resource share. Setting it to '-1' tells the host to follow the resource share set at the project. Any other value sets the resource share for that specific host.

I didn't know this simply because I don't see any "-1" on any table, visually, neither on BAM nor in BOINCMgr. MyProjects table as well as all 12 HostProjects tables all have values that are not -1. Maybe -1 is what is stored in the database, but in the UI, I see no -1. Therefore, I had no clue that it was even an allowable entry.


[BOINCstats Willy wrote:
Your hosts have a mixture of '-1' and other values.

I have no clue how that could have happened, since I didn't know to enter -1, or any other value, for that matter, into the HostProjects tables in BAM. I thought the HostProjects tables displayed what the host was sending back to BAM.


[BOINCstats Willy wrote:
So, if host x has the resource share for SETI set to 100 and you change the resource share for SETI to 50 on the My projects page, host x will not follow and remain at 100.

Does this clear up things or is it even more confusing now?

It does explain why my hosts would get some resshare values from the project and some would stay stubbornly at 100, except for the fact that I have no idea how so many projects on so many hosts were set to -1 because I have never ever changed the values in the HostProjects tables. I only ever played with the MyProjects tables ever since I signed up with BAM in 2015. I didn't know anything else was possible.

Now I am scared to go into the HostProjects tables and set stuff to -1, because I want all projects to have the same resshare on every host as described in MyProjects page, but I don't want all my painful manual work on the machines to go away because of some bug...

Also, there are the two other related(?) issues reported in my last two msgs in this thread.

Tuna

[BOINCstats] Willy
 
Forum moderator - Administrator - Developer - Tester - Translator
BAM!ID: 1
Joined: 2006-01-09
Posts: 9136
Credits: 348,241,820
World-rank: 2,972

2017-04-15 08:52:28

The default value when a host is created is '-1'. If you hover over de input field there is a popup which explains it.
Please do not PM, IM or email me for support (they will go unread/ignored). Use the forum for support.
Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-04-15 09:06:03

Oh dear goodness... One has to click on it, then hover to see that... And, since there is no "FillDown", I have to do that like 55 times per host. And, ENTER key doesn't work to go to the next row after typing -1, so it is CLICK -1 CLICK -1 .... 55 times, for each of the 12 hosts, thus 55*12=660 times CLICK -1.

The worst offenders (those with the most projects where the value was NOT coming from the project and was stuck at 100 in the HostProjects table) were the newer machines I had set up, and they would have gotten their initial projects via the AttachNewHostByDefault=YES setting on MyProjects table, not by me manually setting them per project on that new host. Are you sure there is no bug in that code that puts the current actual resshare value instead of -1 while a new host is doing its first AcctMgrSync and getting its first project list along with all the resshare values?
Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-04-15 22:15:24
last modified: 2017-04-15 22:25:03

Another thought: Even if the you create the new host/project relationships as resshare=-1 (and I'd suggest that they are shown as -1, to indicate that visually), do you think maybe clicking the checkboxes in the project row in the host page ends up not only (un)setting the flags, but also make the value currently shown in the ResShare cell replace the internal -1? After all, clicking on any checkbox also turns that value into an editbox, and maybe they report that they have been edited. That would explain a lot as to why almost all my newer machines were acting as if their individual resshare values were NOT -1, because I played with those checkboxes a lot. Yet another reason to show "-1" if the cell is "use project value", instead of showing an actual value.

Furthermore, showing -1 instead of the default value from the MyProjects page also avoid this problem: I just manually changed everything to -1 for one of my hosts, did a few AcctMgrSync and project updates in BOINCMgr, so all the projects are using the resshare values as defined by the projects. However, that doesn't mean that they are the same as in MyProjects page. For instance, my PrimeGrid setting on MyProjects is 200, the setting is -1 for this host, but the resshare value on PrimeGrid itself is 100, so the host actually uses 100, but the host page on BAM shows 200 because (presumably) it gets that value from the MyProjects table. If it showed -1, that basically tells the user "Go look at the project".

Tuna
[BOINCstats] Willy
 
Forum moderator - Administrator - Developer - Tester - Translator
BAM!ID: 1
Joined: 2006-01-09
Posts: 9136
Credits: 348,241,820
World-rank: 2,972

2017-04-22 13:17:27

Made some changes.
Please do not PM, IM or email me for support (they will go unread/ignored). Use the forum for support.
Tuna Ertemalp
    Donator
Tester
BAM!ID: 37744
Joined: 2007-10-31
Posts: 531
Credits: 9,547,425,088
World-rank: 273

2017-04-22 17:15:57

Awesommmmmmm.....
Pages: [1]

Index :: BAM! Bug Report :: Resource Share issues between BAM & hosts, and hosts & projects
Reason: