Posts by Ananas

1) Message boards : Projects : ABC@home - could someone please help ? (Message 46567)
Posted 3 Dec 2012 by Profile Ananas
Post:
any ideas ? :-)

http://abcathome.com/forum_thread.php?id=891

Yes, but only a small chance though :

Check your browser for a cookie named "auth" for the ABC server - even if you login with mail+password, the browser creates that cookie containing your account ID.
2) Message boards : Teams : Users to be deleted (Message 31013)
Posted 13 Feb 2010 by Profile Ananas
Post:
http://boinc.berkeley.edu/teams/view_profile.php?userid=

1690 up to 1697

I'm not sure wether they are part of the Global Teams import - but they use the profile in forum spam, not so good for Berkeley.
3) Message boards : Questions and problems : Boinc bleibt beim Aufbau einer Verbindung haengen, auch nach Neustart (Message 28099)
Posted 18 Oct 2009 by Profile Ananas
Post:
Stimmt, da hast Du natuerlich Recht. DNS-Zugriffe haengen den Coreclient unter Windows sogar so extrem auf, dass WUs deswegen abbrechen (der Heartbeat-Konzeptfehler)
4) Message boards : Questions and problems : Boinc bleibt beim Aufbau einer Verbindung haengen, auch nach Neustart (Message 28085)
Posted 18 Oct 2009 by Profile Ananas
Post:
Eigentlich hat die Sache mit dem Network-Request mit dem Problem gar nichts zu tun.

Die Fehlermeldung bezieht sich auf die Verbindung zwischen BOINC-Manager und Coreclient lokal oder ueber's Intranet.

Die Netzwerk-Zugriffseinstellung bezieht sich auf die Verbindung zwischen Coreclient und den Projekten ueber's Internet.


Das Problem muss also duch die Neuinstallation weggegangen sein. Moeglicherweise hatte es 'was mit BOINC im Autostart und/oder Service-/Screensaverinstallation zu tun und der Coreclient lief dadurch ebenfalls zweimal. Da er nur auf zwei Ports reagiert, ist der dritte BOINC-Manager "leer ausgegangen" und konnte keine Verbindung mehr herstellen.


Du scheinst mit dem Problem uebrigens nicht alleine zu sein : forum_thread.php?id=4366 bezieht sich auf die gleiche BOINC-Version.

Haengt vielleicht mit einem Bug im neuen Installer zusammen.
5) Message boards : Questions and problems : Question about RAM (Message 28084)
Posted 18 Oct 2009 by Profile Ananas
Post:
Maybe an option :

Put 2 * 512MB into one box, 4 * 256MB into the second and the remaining sticks into the third.

You will not loose Dual Channel on all boxes then.
6) Message boards : Questions and problems : My error in attaching to a project: how to correct? (Message 28083)
Posted 18 Oct 2009 by Profile Ananas
Post:
...
15/10/2009 12:29:57 Cosmology@Home [error] File params_092809_220101_2.ini has wrong size: expected 1911, got 0
...


That's a known bug in the Cosmo project (server side!) and has nothing to do with the BOINC client version.

It happens with all resent workunits. The second (and each following) delivered result is damaged.

Unfortunately each error result and each result not returned in time produces 4 results with download failure, so sometimes you get several download failures in a row.

Just retry until you get a "fresh" result :-)
7) Message boards : Questions and problems : TERRIBLE WAY TO MAKE PROFILES (Message 28081)
Posted 18 Oct 2009 by Profile Ananas
Post:
... You can always write your profile in Notepad and just copy it over.


The clipboard can help too.
8) Message boards : BOINC client : rsc_memory_bound still in effect? (Message 27440)
Posted 16 Sep 2009 by Profile Ananas
Post:
Not only related to the core client ...

Does "<rsc_memory_bound>" still do something?

It's used by the scheduling-server when deciding if host can handle the task or not. If wu.rsc_memory_bound > usable_ram, it won't be sent to computer, there usable_ram is the max of users active and idle memory-limits, or computers installed memory in case of really old client.


Thanks :-)
9) Message boards : BOINC client : rsc_memory_bound still in effect? (Message 27438)
Posted 16 Sep 2009 by Profile Ananas
Post:
Did it ever do anything?


Afaik. there have been client versions that checked it, because in some projects people complained about it.

It might have been the global settings / venue RAM limit as well, so they just assumed that it has been the WU RAM limit.

If it turns out that it never has been used, it should just silently disappear from program codes, database and especially from the documentation. Changing a placebo in order to fix a bug can lead the project guys to wrong conclusions when trying to fix problems.

edit :

4.11 did use it.

int ACTIVE_TASK::init(RESULT* rp) {
    result = rp;
    wup = rp->wup;
    app_version = wup->avp;
    max_cpu_time = rp->wup->rsc_fpops_bound/gstate.host_info.p_fpops;
    max_disk_usage = rp->wup->rsc_disk_bound;
    max_mem_usage = rp->wup->rsc_memory_bound;

    strcpy(process_control_queue.name, rp->name);
    strcpy(graphics_request_queue.name, rp->name);

    return 0;
}

...

#if 0
// if an app has exceeded its maximum allowed memory, abort it
//
bool ACTIVE_TASK::check_max_mem_exceeded() {
    // TODO: calculate working set size elsewhere
    if (working_set_size > max_mem_usage || working_set_size/1048576 > gstate.global_prefs.max_memory_mbytes) {
        msg_printf(
            result->project, MSG_INFO,
            "Aborting result %s: exceeded memory limit %f\n",
            result->name,
            min(max_mem_usage, gstate.global_prefs.max_memory_mbytes*1048576)
        );
        abort_task("Maximum memory usage exceeded");
        return true;
    }
    return false;
}
#endif

bool ACTIVE_TASK::check_max_mem_exceeded() {
    if (max_mem_usage != 0 && rss_bytes > max_mem_usage) {
        msg_printf(
            result->project, MSG_INFO,
            "result %s: memory usage %f exceeds limit %f\n",
            result->name,
            rss_bytes,
            max_mem_usage
        );
        abort_task("Maximum memory usage exceeded");
        return true;
    }
    return false;
}


The "abort_task" part has been commented out in almost every version - but the code dated Sep. 24 2004 (must be 4.11) has it active.
10) Message boards : BOINC client : rsc_memory_bound still in effect? (Message 27435)
Posted 16 Sep 2009 by Profile Ananas
Post:
Not only related to the core client ...

Does "<rsc_memory_bound>" still do something?

It is still documented but I cannot find any hint that it is still used somewhere - besides beeing converted into a numeric value now and then and traveling through several program modules (client, server and utils), sometimes cloaked as "max_mem_usage".

The memory limits that still seem to be checked are the ones from the global settings.


Of course it needs to be there for old clients - but if it isn't used for anything, it could as well be ignored/skipped while parsing the result XML data by later clients.
11) Message boards : Projects : Fake BOINC project pages (Attn., project stalkers) (Message 26659)
Posted 18 Aug 2009 by Profile Ananas
Post:
Lately a lot of web pages pop up having "well-known" keywords on them that usually occur only on BOINC project pages in this combination.

Project hunters need to be careful not to click those, they will kidnap your browser to criminal fake virus scanner sites (search engine spiders do not see that!).

Those pages usually do not belong to the web site they are hosted on, I suppose they have been installed through a root kit or something similar.

Paying attention to the URL, they are recongnizable, e.g. :

(base directory)/uwfep/mllal/participants.php

with this Google excerpt : "returning participants view stats modify preferences organizers feud and participants finding focus group participants in uk armenia genocide participants"

The URL pattern used to be a 4/4 random letter combination, they switched to 5/5 lately after Google has been informed about the problem. I have seen a few with just a single random 7-letter directory. This means that they modify this URL pattern now and then!


I guess, the collections / combinations of keywords come from meta search engines that have one single purpose : collecting commonly used search keywords.
12) Message boards : Projects : Validate WinXP vs. Win2000 under MinGW (Message 26656)
Posted 18 Aug 2009 by Profile Ananas
Post:
We found that there are small calculation differences in the versions of msvcrt.dll, that come with Windows 2000 based and Windows XP based operating systems.

One possible solution is to use a specific VC runtime by overriding the default DLL in the linker specs, e.g. :

CC       = mingw32-gcc ..... -specs=local_gccspecs
...
all: 	local_gccspecs $(PROGS)

local_gccspecs: .
<tab>mingw32-gcc -dumpspecs|sed 's/-lmsvcrt/-lmsvcr71/g'>local_gccspecs


This worked for us in the 32-bit 4.4.0 of MinGW/GCC, it will probably not work with the x64 4.5.0 because there seems not to be such a DLL there (and thus no import library either of course)

I assume that the functions that are affected are the exponential ones.
13) Message boards : Server programs : Is that a BOINC server virus? (Message 26114)
Posted 19 Jul 2009 by Profile Ananas
Post:
This might sound stupid - but have you checked the space on the file systems (including the /tmp one)?

Some programs that do not catch write errors tend to go into extremely time consuming loops.
14) Message boards : Server programs : Daily quota issues (Message 26112)
Posted 19 Jul 2009 by Profile Ananas
Post:
Currently "didn't need", "redundant result", "cancelled" and stuff like that reduce the daily quota of the hosts, server side issues shouldn't influence the daily quota.

==========================================

The current quota systems allows hosts to always have the full quota with only 2% (!!!) good results :

void SCHEDULER_REPLY::got_good_result() {
host.max_results_day *= 2;
if (host.max_results_day > config.daily_result_quota) {
host.max_results_day = config.daily_result_quota;
}
}

void SCHEDULER_REPLY::got_bad_result() {
host.max_results_day -= 1;
if (host.max_results_day < 1) {
host.max_results_day = 1;
}
}

This allows crunchers to choose the more "lucrative" results and abort the ones with a lower credits/time ratio.

The bad-hosts penalty should be way more efficient :

void SCHEDULER_REPLY::got_good_result() {
host.max_results_day += 1;
if (host.max_results_day > config.daily_result_quota) {
host.max_results_day = config.daily_result_quota;
}
}

void SCHEDULER_REPLY::got_bad_result() {
host.max_results_day /= 2;
if (host.max_results_day < 1) {
host.max_results_day = 1;
}
}

==========================================

Those two issues should be solved in the same release. Some projects cancel large batches of results frequently so improving the bad-hosts issue requires that server-side issues cannot make a host "bad".


p.s.: I did report the 2% problem already somewhere long time ago but I didn't find where it was
15) Message boards : Server programs : Server-side DCF / Duration Correction Factor (Message 26111)
Posted 19 Jul 2009 by Profile Ananas
Post:
Thanks for the info, I didn't know that.
16) Message boards : Server programs : Server-side DCF / Duration Correction Factor (Message 25972)
Posted 12 Jul 2009 by Profile Ananas
Post:
The rsc_fpops_est values of some projects are way off.

It would sure be a good option for the project admins, if there was a feedback like this :

effective_rsc_fpops_est = SELECT workunit.rsc_fpops_est * AVG (host.duration_correction_factor) WHERE host.expavg_credit > 100

well, something like this, where effective_rsc_fpops_est would be what the scheduler actually uses for result delivery.


The AVG expression needs to be cached somewhere (maybe on a weekly base) of course instead of computing it for each result.
17) Message boards : BOINC client : Heartbeat replacement (Message 25573)
Posted 21 Jun 2009 by Profile Ananas
Post:
I think I know our enemy now, it's one (or both) of those two :

svchost.exe -k NetworkService

svchost.exe -k netsvcs

The same occurs on my PC at work, not with BOINC there but other networking applications become unresponsive when one of these two takes 99% CPU load.

Having multiple CPUs seems not to help, they seem to slow down some core services that affect all TCP applications.


p.s.: As the API communicates through memory, not local TCP, it will not help to use CPU seconds instead of realtime seconds for the heartbeat timeout :-(
18) Message boards : Web interfaces : My account or your account? (Message 25161)
Posted 1 Jun 2009 by Profile Ananas
Post:
You're right about the BOINC manager buttons (maybe some project guys will read this) - but the gui_urls.xml doc says :

    <gui_url>
        <name>Your account</name>
        <description>View your account information and credit totals</description>
        <url>http://foo.project.com/blah.php?userid=<userid></url>
    </gui_url>



Basically you're right when it comes to the project pages as well ... but ...

In "user/sample_index.php" the link to "home.php" is labeled "Your account"

In "project.sample/project.inc" the link to "home.php" is labeled "My account"

So this is where the three versions originate from and that's what all projects start with, they copy those samples, adjust what they need and soon they have "won" those inconsistencies (sp?)

As soon as the template and docs handle it consistent, at least all future projects will have a better chance to get this right.
19) Message boards : Web interfaces : My account or your account? (Message 25153)
Posted 1 Jun 2009 by Profile Ananas
Post:
I hope this isn't considered to be knitpicking ...

It is when you spell it like that. Loose the 'k'. :-)

I am nitpicking on you there... ;-)


I would have spelled it like that (without the 'k') but wasn't sure - so looked on the web and found both spellings - should have asked GoogleFight instead ;-)

edit : Oh, now I see my "problem" ... I searched for "nittpicking" and it offered me : Did you mean: knitpicking
20) Message boards : Web interfaces : My account or your account? (Message 25151)
Posted 1 Jun 2009 by Profile Ananas
Post:
I hope this isn't considered to be knitpicking ...

The phrases "Your account" and "My account" are used a bit randomly, in the bottom line it's "Your", on the front pages "My" and in the additional project links (BOINC manager) both can be found too.

Some of those additional BM (My or Your account) links lead to "home.php" (OK), others to "show_user.php" (which doesn't make so much sense).

Of course everyone knows that "My account" is "Your account" (well, if you have my account key that is *grin*) - but I still think it should be fixed sometimes.


Next 20

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.