Classical run OK, trajtou crash asap


advanced search

Message boards : Number crunching : Classical run OK, trajtou crash asap

Reply to this thread
Subscribe to this thread
Sort
AuthorMessage
Saenger User profile image
Avatar
private message
Joined: Feb 15, 2006
Posts: 18
ID: 179
Credit: 27,250
RAC: 8
Message 2423 - Posted 13 Jun 2009 12:05:54 UTC
Last modified: 13 Jun 2009 12:09:43 UTC

Test test test, this is no spam (wtf does akismet want? no links to problems?)

Edit (to insert the real info):
I've received a lot of WUs lately, nearly all of them crashed asap. All of the crashed ones are trajtou, while Classical work fine.

Next edit with some links:
My WUs
A trajtou with a lot of errors
____________
Gruesse vom Saenger


For questions about Boinc look in the BOINC-Wiki

Saenger User profile image
Avatar
private message
Joined: Feb 15, 2006
Posts: 18
ID: 179
Credit: 27,250
RAC: 8
Message 2424 - Posted 13 Jun 2009 13:06:18 UTC
Last modified: 13 Jun 2009 13:07:30 UTC

F***ing hell!
What's this bloody akismet doing here?

Unable to handle request

Your post has been marked as spam by akismet.net anti-spam system. If you feel that this is wrong, please try editing your message.

I'll try again with my next post.

Edit:
Editing is of course not possible, as it's nowhere stored. It's "write the post again from scratch".

Saenger User profile image
Avatar
private message
Joined: Feb 15, 2006
Posts: 18
ID: 179
Credit: 27,250
RAC: 8
Message 2425 - Posted 13 Jun 2009 13:07:56 UTC
Last modified: 13 Jun 2009 13:10:29 UTC

Now for the real post, content only via edit possible.

Edit for insertion of content:
Why do I even get those trajtous? A look at the applications page shows no app available for my penguin.
What goes wrong with the scheduler?

ritterm User profile image
private message
Joined: Jun 19, 2008
Posts: 4
ID: 15306
Credit: 125,116
RAC: 329
Message 2426 - Posted 13 Jun 2009 16:41:27 UTC
Last modified: 13 Jun 2009 16:47:05 UTC

I'm having similar problems with the trajtou jobs. My Windows clients are doing fine with these WUs, but those running on my 64-bit Linux host (C2Q running Ubuntu 9.04 and BOINC mgr 6.2.18) keep failing. At first they did so immediately and I found that I needed to add a couple of libraries. Since then they run for about 30 minutes before failing. The latest show the following stderr out:

<core_client_version>6.2.18</core_client_version>
<![CDATA[
<message>
process exited with code 64 (0x40, -192)
</message>
<stderr_txt>
</stderr_txt>
]]>

The result IDs are 14625531 and 14625610.

Jeff Gu
Avatar
private message
Joined: Apr 9, 2006
Posts: 10
ID: 1065
Credit: 892,680
RAC: 89
Message 2427 - Posted 13 Jun 2009 19:27:50 UTC

I'm having the same problem on all of the Linux machines with the trajtou work units. They error out immediately. I have both the libstdc++5 and libstdc++6 libraries installed. The machines are all running either Xubuntu or Ubuntu versions 8.04, 8.10 or 9.04 and all versions seem to be having the same problem.
____________
Jeff Gu
Guru Mountain DC Team

Jeff Gu
Avatar
private message
Joined: Apr 9, 2006
Posts: 10
ID: 1065
Credit: 892,680
RAC: 89
Message 2428 - Posted 14 Jun 2009 2:55:02 UTC

Saenger seems to have found the best clue. No Linux 64 application is listed for trajtou... only Windows... so I'm guessing he's right about the scheduler being hosed...


____________
Jeff Gu
Guru Mountain DC Team

Saenger User profile image
Avatar
private message
Joined: Feb 15, 2006
Posts: 18
ID: 179
Credit: 27,250
RAC: 8
Message 2429 - Posted 14 Jun 2009 6:15:47 UTC - in response to Message ID 2428.
Last modified: 14 Jun 2009 6:28:52 UTC

Saenger seems to have found the best clue. No Linux 64 application is listed for trajtou... only Windows... so I'm guessing he's right about the scheduler being hosed...

I've found some strange files in my /projects/boinc.gorlaeus.net folder:
    [*] classical_5.56_x86_64-pc-linux-gnu.exe
    [*] trajtou-cu111_5.37_x86_64-pc-linux-gnu.exe
    [*] trajtou-pd110paw_5.37_x86_64-pc-linux-gnu.exe


They are as well here in the download area: http://boinc.gorlaeus.net/download/

I dunno why they are .exe, but at least the first one seems to run nevertheless.

After installing the c++5 library mine run as well for some time before crashing, so I had to put this project on NNW until I got a sufficient answer here.

The new stderr is this:

<core_client_version>6.4.5</core_client_version>
<![CDATA[
<stderr_txt>
Unrecognized XML in parse_init_data_file: computation_deadline
Skipping: 1245469502.000000
Skipping: /computation_deadline
Unrecognized XML in GLOBAL_PREFS::parse_override: mod_time
Skipping: /mod_time
Unrecognized XML in GLOBAL_PREFS::parse_override: max_ncpus_pct
Skipping: 100.000000
Skipping: /max_ncpus_pct

</stderr_txt>
]]>

____________
Gruesse vom Saenger


For questions about Boinc look in the BOINC-Wiki

m.somers User profile image
Forum moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Avatar
private message
Joined: Nov 14, 2005
Posts: 630
ID: 1
Credit: 1,417,572
RAC: 2
Message 2431 - Posted 14 Jun 2009 21:56:04 UTC

There was an issue with the linux trajtou apps when I needed to clean the db (see news item) because of some limit in the backend. Fixed a few days ago already but WUs were send out and most of them succeeded.

m.
____________
M.F. Somers

Skip Da Shu
Avatar
private message
Joined: Apr 29, 2007
Posts: 23
ID: 5991
Credit: 855,139
RAC: 62
Message 2433 - Posted 15 Jun 2009 3:28:56 UTC - in response to Message ID 2431.
Last modified: 15 Jun 2009 3:43:06 UTC

There was an issue with the linux trajtou apps when I needed to clean the db (see news item) because of some limit in the backend. Fixed a few days ago already but WUs were send out and most of them succeeded.

m.


So what's the deal with the APPs page then? Just out of date? Or are you saying the dB error caused Windoze apps to go to Linux clients? I'd better go look for news item...

____________
- da shu @ HeliOS,
"La carencia de recursos no debe de ser impedimento para que un niño tenga acceso a la tecnología."

Jeff Gu
Avatar
private message
Joined: Apr 9, 2006
Posts: 10
ID: 1065
Credit: 892,680
RAC: 89
Message 2434 - Posted 15 Jun 2009 3:43:50 UTC

I'm still having trajtou work units on the Linux machines ending as computation errors anywhere from a few seconds to many hours into the WU... with 20 machines running this project, that adds up to an awful lot of days of wasted electricity that could have been better spent on another project. I'm now having to go through every machine and abort all of the trajtou work units. Some of the machines were tied up for days with these work units running in high priority mode, only to fail to meet the deadline, anyway, or to error out.

I'm rather proud of the fact that a team the size of ours made it to a 15th place world ranking with this project. Personally, I was impressed when Docking@Home sent out an email immediately when a batch of work units were bad, and getting stuck... I got the email before I'd wasted more than two hours on them. I wish other projects, including this one, could find the time to be this courteous to the volunteers that make their project possible.
____________
Jeff Gu
Guru Mountain DC Team

m.somers User profile image
Forum moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Avatar
private message
Joined: Nov 14, 2005
Posts: 630
ID: 1
Credit: 1,417,572
RAC: 2
Message 2437 - Posted 15 Jun 2009 9:41:34 UTC

There shouldn't be any; the linux trajtou application was removed as soon as I saw some of the WUs crash. Somehow, the linux hosts keep on requesting work for the trajtou apps even though the app was removed from the DB last week. Anyway, I forced an app version update today on the trajtou apps to clear the hosts. Let's wait and see if this clears things up...

m.
____________
M.F. Somers

Reply to this thread

Message boards : Number crunching : Classical run OK, trajtou crash asap



Return to Leiden Classical main page


Copyright © 2013 Leiden University - Leiden Institute of Chemistry - Theoretical Chemistry Department