-return_results_immediately


log in

Advanced search

Questions and Answers : Unix/Linux : -return_results_immediately

Author Message
Tixomir
Send message
Joined: 5 Oct 02
Posts: 2
Credit: 286
RAC: 0
Serbia
Message 1812 - Posted: 25 Jun 2004, 10:46:52 UTC

i use dial-up to conecto to internet, and when seti finishes one data file, it goes to another, and tryes to send done material, but if I'm offline it fails, later, i connect, ad try -return_results_immediately and ther's no network transfer, seti continues with it's new job... some ideas?

Profile Doris and Jens
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 21 Nov 99
Posts: 362
Credit: 1,029,545
RAC: 340
Germany
Message 1944 - Posted: 25 Jun 2004, 15:28:37 UTC
Last modified: 25 Jun 2004, 15:29:04 UTC

There are two steps after you finish a result.

1. BOINC try to upload the result file to the data server
2. BOINC reports the successfully uploaded result to the scheduler

Step 1. happens immediately after the WU is crunched and may fail because of missing internet connection. Then BOINC set a deleay (increasing up to 4 hours) and retry after this deleay time. On windows dialup hosts BOINC asked (if customized) befor connect to the internet, don't know if this works on linux. Have no dialup linux.

Step 2. happens normally when the client ask the scheduler for new work. Because this can be to long to wait for some users, they use the option -return_results_immediately. This means BOINC didn't wait until there is need for new work, it connect the scheduler after each upload. In other case BOINC is collecting the reports about the uploads.

You can force BOINC to connect the scheduler manually with the option: -update prefs [URL] where URL is the projekt URL.

Greetings from Bremen/Germany

Jens Seidler



Tixomir
Send message
Joined: 5 Oct 02
Posts: 2
Credit: 286
RAC: 0
Serbia
Message 1982 - Posted: 25 Jun 2004, 17:09:18 UTC - in response to Message 1944.
Last modified: 25 Jun 2004, 17:11:16 UTC

Thanks for your answer, now i understand how Boinc handles resulting data, just one question to make this clear.
There is no way to force sending of data (deleayed step 1) like I can force update of scheduler with -update prefs [URL]?


P.S. I can set linux to connect to internet when Bionic need's it, but I don't want that solution :/

Profile Doris and Jens
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 21 Nov 99
Posts: 362
Credit: 1,029,545
RAC: 340
Germany
Message 2279 - Posted: 26 Jun 2004, 5:57:21 UTC - in response to Message 1982.
Last modified: 26 Jun 2004, 5:58:09 UTC

> There is no way to force sending of data (deleayed step 1) like I can force
> update of scheduler with -update prefs [URL]?

On Window this is build in the GUI. As I remember the CLI client is missing this option at the moment. I agree that it is needed, but think it would be come at first with the new Linux (universal) GUI. I hope that it is finshed this year.

Greetings from Bremen/Germany

Jens Seidler (TheBigJens)



Profile Michael MacKay
Send message
Joined: 22 Jul 99
Posts: 2
Credit: 9,945,749
RAC: 5,814
Canada
Message 6211 - Posted: 10 Jul 2004, 19:02:21 UTC - in response to Message 1944.

> Step 1. happens immediately after the WU is crunched and may fail because of
> missing internet connection. Then BOINC set a deleay (increasing up to 4
> hours) and retry after this deleay time. On windows dialup hosts BOINC asked
> (if customized) befor connect to the internet, don't know if this works on
> linux. Have no dialup linux.
>
> Step 2. happens normally when the client ask the scheduler for new work.
> Because this can be to long to wait for some users, they use the option
> -return_results_immediately. This means BOINC didn't wait until there
> is need for new work, it connect the scheduler after each upload. In other
> case BOINC is collecting the reports about the uploads.
>
> You can force BOINC to connect the scheduler manually with the option:
> -update prefs [URL] where URL is the projekt URL.
>

This doesn't do anything.

On a Solaris box doing dial-up PPP:

# ./boinc_3.18_sparc-sun-solaris2.7 -update_prefs http://setiathome.berkeley.edu
2004-07-10 14:54:37 [---] Starting BOINC client version 3.18 for sparc-sun-solaris2.7
2004-07-10 14:54:37 [SETI@home] Project prefs: no separate prefs for home; using your defaults
2004-07-10 14:54:37 [SETI@home] Host ID is 61100
2004-07-10 14:54:37 [---] General prefs: from SETI@home (last modified 2004-07-09 14:03:29)
2004-07-10 14:54:37 [---] General prefs: using separate prefs for home
2004-07-10 14:54:37 [SETI@home] Sending request to scheduler: http://setiboincdata.ssl.berkeley.edu/sah_cgi/cgi
2004-07-10 14:54:46 [SETI@home] Scheduler RPC to http://setiboincdata.ssl.berkeley.edu/sah_cgi/cgi succeeded
2004-07-10 14:54:46 [SETI@home] Project prefs: no separate prefs for home; using your defaults

And at this point the output hangs indefinitely, and there is no network
traffic. So how do I connect to the SETI@home server when I'm actually
connected to the Internet, and not according to some mystery schedule
determined by the BOINC software?

There must be many participants, like me, who will make a dial-up connection
because we have finised work to upload. But BOINC doesn't give us the option:
by the time we notice the work is ready, it is too late. Since we're not
connected, we get:

Can't resolve hostname setiboincdata.ssl.berkeley.edu (host not found)

and then something like:

Backing off 1 hours, 8 minutes, and 35 seconds on transfer of file setiathome_3.08_sparc-sun-solaris2.7

Tough luck for us, I guess, for not being on-line at the correct second!

Disappointed...

Michael

phunks
Send message
Joined: 26 Nov 02
Posts: 1
Credit: 1,373,439
RAC: 901
Canada
Message 10812 - Posted: 21 Jul 2004, 16:18:30 UTC


There is a way:

1/ shut down BOINC

2/ open client_state.xml in an editor

3/ find these tags: next request time and min_rpc_time

4/ replace the values in both places with 0 ## that's a zero

5/ save the file and restart Boinc with -update_prefs URL

That's it.

It's a little more work but more enjoyable than waiting out a "3 hour, 10
minute, 28 seconds" deferment notice.

Paul

Questions and Answers : Unix/Linux : -return_results_immediately

Copyright © 2014 University of California