Message boards : Questions and problems : Can't get BOINC to use proxy
Message board moderation
Author | Message |
---|---|
Send message Joined: 12 Jun 08 Posts: 7 |
Got a new laptop (XP professional, SP2). Installed BOINC 5.10.45 - so far so good. Tried to attach to project - received "project temporarily unavailable". Tried another project with same results. From previous experience I know that this might be caused by BOINC not being able to pass through the autoproxy setup my company uses; however, no proxy configuration form was presented to me. No problem, went to the advanced->options->http proxy form and filled the details - but I'm unable to store them because this form doesn't have an OK button in this installation! Tried modifying the client_state.xml to add the proxy data. This helped with getting the data in, but this doesn't check the "use proxy" button - either I need the form to do this or there's an option I can add in the XML but can't see it documented anywhere. What gives? |
Send message Joined: 29 Aug 05 Posts: 15542 |
went to the advanced->options->http proxy form and filled the details - but I'm unable to store them because this form doesn't have an OK button in this installation! What size of screen resolution do you use? Might be that the OK button falls off the screen. The OK button is available all the time on the main window. Else press Enter. What you can do is add the <force_auth> flag option in cc_config.xml See this FAQ on how to set up cc_config.xml You only need these entries: <cc_config> <options> <force_auth>N</force_auth> </options> </cc_config> Where N is either basic or digest or gss-negotiate or ntlm |
Send message Joined: 12 Jun 08 Posts: 7 |
Thanks, using 'enter' solved this problem... Stupid of me not to have thought to try it. I'm using 1280X1024, should be more than enough... I see the entire frame of the form but it seems to be too large for the frame. In the "connections" dialogue, for example, the left sides of "set default" and "clear default" buttons are placed over the scrollbar of the list, and only the "default connection" header below can be seen but not the default connection itself. Now another question: once I attached to my project (climateprediction.net) I got a task and it seems to run fine. However, the system is dual-core and I'd like to get two tasks - but I'm not getting a second. The manager is sending requests for more work but the server refuses to send any - why would that happen? Thanks, Micha. |
Send message Joined: 29 Aug 05 Posts: 15542 |
The manager is sending requests for more work but the server refuses to send any - why would that happen? Can you post the messages in question, please? Just copy them from the messages tab and post them here. Around 20 of them. |
Send message Joined: 12 Jun 08 Posts: 7 |
The manager is sending requests for more work but the server refuses to send any - why would that happen? 6/12/2008 5:07:27 PM|climateprediction.net|Message from server: No work sent 6/12/2008 5:08:32 PM|climateprediction.net|Sending scheduler request: To fetch work. Requesting 108000 seconds of work, reporting 0 completed tasks 6/12/2008 5:08:37 PM|climateprediction.net|Scheduler request succeeded: got 0 new tasks 6/12/2008 5:08:37 PM|climateprediction.net|Message from server: No work sent 6/12/2008 5:09:40 PM|climateprediction.net|Sending scheduler request: To fetch work. Requesting 108000 seconds of work, reporting 0 completed tasks 6/12/2008 5:09:46 PM|climateprediction.net|Scheduler request succeeded: got 0 new tasks 6/12/2008 5:09:46 PM|climateprediction.net|Message from server: No work sent 6/12/2008 5:11:20 PM|climateprediction.net|Sending scheduler request: To fetch work. Requesting 108000 seconds of work, reporting 0 completed tasks 6/12/2008 5:11:25 PM|climateprediction.net|Scheduler request succeeded: got 0 new tasks 6/12/2008 5:11:25 PM|climateprediction.net|Message from server: No work sent I can post more but they're similar to these. Thanks, Micha. |
Send message Joined: 29 Aug 05 Posts: 15542 |
Ok, let's see what your system does then. By default BOINC only shows messages that aren't telling much, so not to scare people. ;-) If you have the cc_config.xml file already, as I showed how to make above, please edit it and put in: <cc_config> <log_flags> <http_debug>1</http_debug> <proxy_debug>1<proxy_debug> <sched_op_debug>1<sched_op_debug> </log_flags> <options> <force_auth>N</force_auth> </options> </cc_config> To make BOINC read it without restarting, open BOINC Manager->(Advanced view->)Advanced->Read config file. Then make communications again and repost the messages please. |
Send message Joined: 12 Jun 08 Posts: 7 |
Then make communications again and repost the messages please. Well, there you go (proxy name and IP address deleted since they belong to my company): 6/12/2008 10:50:50 PM|climateprediction.net|[sched_op_debug] Starting scheduler request 6/12/2008 10:50:50 PM|climateprediction.net|Sending scheduler request: To fetch work. Requesting 108000 seconds of work, reporting 0 completed tasks 6/12/2008 10:50:50 PM||[http_debug] HTTP_OP::init_post(): http://climateapps2.oucs.ox.ac.uk/cpdnboinc_cgi/cgi 6/12/2008 10:50:50 PM||[http_debug] [ID#0] info: About to connect() to proxy proxy01....com port 911 (#0) 6/12/2008 10:50:50 PM||[http_debug] [ID#0] info: Trying 14...76.35... 6/12/2008 10:50:50 PM||[http_debug] [ID#0] info: Connected to proxy01....com (14...76.35) port 911 (#0) 6/12/2008 10:50:50 PM||[http_debug] [ID#0] Sent header to server: POST http://climateapps2.oucs.ox.ac.uk/cpdnboinc_cgi/cgi HTTP/1.1 User-Agent: BOINC client (windows_intelx86 5.10.45) Host: climateapps2.oucs.ox.ac.uk Pragma: no-cache Accept: */* Accept-Encoding: deflate, gzip Proxy-Connection: Keep-Alive Content-Type: application/x-www-form-urlencoded Content-Length: 6350 Expect: 100-continue 6/12/2008 10:50:52 PM||[http_debug] [ID#0] info: HTTP 1.0, assume close after body 6/12/2008 10:50:52 PM||[http_debug] [ID#0] Received header from server: HTTP/1.0 200 OK 6/12/2008 10:50:52 PM||[http_debug] [ID#0] Received header from server: Date: Thu, 12 Jun 2008 19:50:50 GMT 6/12/2008 10:50:52 PM||[http_debug] [ID#0] Received header from server: Server: Apache/2.2.3 (Unix) PHP/5.1.6 6/12/2008 10:50:52 PM||[http_debug] [ID#0] Received header from server: Content-Type: text/xml 6/12/2008 10:50:52 PM||[http_debug] [ID#0] Received header from server: X-Cache: MISS from hafwpr01....com 6/12/2008 10:50:52 PM||[http_debug] [ID#0] Received header from server: X-Cache-Lookup: MISS from hafwpr01....com:911 6/12/2008 10:50:52 PM||[http_debug] [ID#0] Received header from server: Via: 1.0 hafwpr01....com:911 (squid/2.6.STABLE18) 6/12/2008 10:50:52 PM||[http_debug] [ID#0] Received header from server: Proxy-Connection: close 6/12/2008 10:50:52 PM||[http_debug] [ID#0] Received header from server: 6/12/2008 10:50:52 PM||[http_debug] [ID#0] info: Closing connection #0 6/12/2008 10:50:55 PM|climateprediction.net|Scheduler request succeeded: got 0 new tasks 6/12/2008 10:50:55 PM|climateprediction.net|[sched_ops_debug] Server version 603 6/12/2008 10:50:55 PM|climateprediction.net|Message from server: No work sent 6/12/2008 10:50:55 PM|climateprediction.net|[sched_op_debug] Deferring communication for 48 min 40 sec 6/12/2008 10:50:55 PM|climateprediction.net|[sched_op_debug] Reason: no work from project |
Send message Joined: 29 Aug 05 Posts: 15542 |
6/12/2008 10:50:55 PM|climateprediction.net|[sched_op_debug] Reason: no work from project That's probably the best answer. I'll wait for confirmation from one of the CPDN people that roam these boards as well. |
Send message Joined: 25 Nov 05 Posts: 1654 |
Yes, the long models, which is the default type, had run out, but more have now been generated. |
Send message Joined: 13 Aug 06 Posts: 778 |
Hi Micha CPDN now has plenty of new work available. But if you've selected only SAP HADAM models in your CPDN project preferences and your computer has less than 1.5GB RAM, the server won't send you anything. That's the default minimum RAM to run those models. You'd need to select HADCM (very long) or HADCM (much shorter) instead. http://climateapps2.oucs.ox.ac.uk/cpdnboinc/server_status.php |
Send message Joined: 12 Jun 08 Posts: 7 |
Hi Micha I have 2GB RAM and haven't selected any preference to a certain type of models. Is it possible that the server avoids sending me more work either because of other preferences (e.g. disk space etc), because it takes into account the resources taken by the model already running on one core or because there's a daily limit on work allocated? Thanks, Micha. |
Send message Joined: 29 Aug 05 Posts: 15542 |
Is it possible that the server avoids sending me more work either because of other preferences (e.g. disk space etc), because it takes into account the resources taken by the model already running on one core or because there's a daily limit on work allocated? If you have disk space problems, BOINC will tell you about that. If you have reached your daily quota, BOINC will tell you about that. So if you don't see any warning messages, try an update on CPDN. |
Send message Joined: 13 Aug 06 Posts: 778 |
Yesterday when no HADCM models were available the CPDN server should have sent Micha one of the two other types instead because he hadn't selected a particular type. Micha, I can't find your name on BoincStats to get to your CPDN account and look for possible clues there. Could you please tell us your CPDN user ID number which will be near the top of your BOINC manager messages. It could be that the BOINC scheduler is refusing to let the computer have an extra model to run CPDN on both cores because of the % resource share Micha has allocated. CPDN's resource share may be too low. In such cases I don't know what BOINC manager message would be generated. I rather think it would just be a 'no work' message with no indication that the resource share allocation for the project has been reached or already exceeded. It used to be the case that if you suspended the model you already have, the CPDN server would send you an extra one, but with newer versions of BOINC I don't think the BOINC scheduler allows this now. |
Send message Joined: 29 Aug 05 Posts: 15542 |
Something everyone seems to miss is that his BOINC was asking for 108,000 seconds of work at the time of not getting work from project (due to no work from project). Requesting 108000 seconds of work, reporting 0 completed tasks The 1.5GB need per model only comes into play when you're actually running two models at the same time. Beforehand BOINC won't know how big a need for memory is needed, nor can the scheduler see if that memory is in use already. All that happens is that BOINC reports to the scheduler that it has 2GB of memory available, which is enough to be able to get work. |
Send message Joined: 13 Aug 06 Posts: 778 |
The 1.5GB need per model only comes into play when you're actually running two models at the same time. The 1.5GB only applies to SAP-type HADAM models. As far as I know it's a minimum requirement for the computer, not a limit per HADAM model. I think a quad with 1.5GB RAM (!!) and HADAM-only selected in CPDN preferences could get 4 of these models from the CPDN server. This quad belonging to Les Bayliss got 4 HADAMs, though it has 3 or 4 GB RAM, not the 4 X 1.5GB = 6GB that would be required if the limit was applied per HADAM. But this restriction isn't relevant to Micha because his computer has 2GB, and his CPDN preferences allow all 3 types of model. Yesterday when there were no HADCM models, the other 2 types were available. If one type of model isn't available the CPDN server sends one of the other types selected. I wonder whether Micha has restricted BOINC or CPDN to using only one core of the computer. Although in that case his BOINC shouldn't be requesting more work. I wonder why Micha's BOINC requested 108000 seconds of work from CPDN. That's 30 hours. There are no models that last 30 hours. |
Send message Joined: 12 Jun 08 Posts: 7 |
Yesterday when no HADCM models were available the CPDN server should have sent Micha one of the two other types instead because he hadn't selected a particular type. Hi, My user ID is 19147 and the computer ID for this machine in the CPDN Web site is 878404. Please let me know if there's any other info you need. Thanks, Micha. |
Send message Joined: 13 Aug 06 Posts: 778 |
Hi Micha Those details are very useful, thank you. I recognise your CPDN name as a longstanding contributor and I see from your CPDN details and from BoincStats that you aren't attached to any other projects. So resource share isn't the problem. Here's the computer. It's listed correctly with 2 cores, 2GB RAM and has a quota of 4 tasks per day. It's already downloaded and started running a HADAM model. I don't see anything visibly wrong there. In this laptop's Boinc Manager Advanced menu Preferences, have you changed any of the default settings? Does the Processor usage tab in that window also show 2 CPUs? Just an idea.... Could a laptop be prevented from downloading a second model/work for the second core because of thermal throttling? |
Send message Joined: 13 Aug 06 Posts: 778 |
I've just tried downloads from CPDN. I selected HADAM only to make the situation similar to Micha. My dual-core with 2Gb RAM, therefore similar to Micha's laptop, has two CPDN Beta models that should complete in about 6 days and my additional work buffer is set to 10 days. No other work from other projects. So I thought that when I set CPDN to Allow new tasks, the CPDN server would send me a model to keep the computer busy during the last 3 days of the work buffer period. But this didn't happen. I clicked Update for CPDN and got 13/06/2008 18:15:35|climateprediction.net|Sending scheduler request: Requested by user. Requesting 0 seconds of work, reporting 0 completed tasks 13/06/2008 18:15:40|climateprediction.net|Scheduler request succeeded: got 0 new tasks I then suspended one of the Beta models and updated CPDN again. The server immediately sent me one HADAM model: 13/06/2008 18:16:56|climateprediction.net|Sending scheduler request: To fetch work. Requesting 872640 seconds of work, reporting 0 completed tasks 13/06/2008 18:17:01|climateprediction.net|Scheduler request succeeded: got 1 new tasks Something doesn't seem to be working properly. You shouldn't need to suspend your current task(s) to get new work to fill the buffer period. I think I've seen this CPDN server behaviour before; I don't think it's the way BOINC is supposed to function with v.5.10.28. Micha, you could try suspending your current model in the Boinc Manager Tasks tab, then updating CPDN. Maybe this will wake up the CPDN server. |
Send message Joined: 25 Nov 05 Posts: 1654 |
I don't think that any of this will solve your problem, but I can't think of anything else. In your Account on the server: 1) Work out which venue you're using for the laptop. (If it says (--), it's the default.) 2) Check the preferences for the venue, and if necessary change them: In Computing preferences, Check that the number of processors is 2 50% of memory when computer is in use 75% of memory when computer is not in use (Those 2 should give the models the best chance of running. Allow them more if possible. It depends on what else the computer is used for.) In climateprediction.net preferences Tick the first model type (HADSM3), and make sure the other 2 AREN'T ticked. In the Projects tab of the manager, click Update |
Send message Joined: 12 Jun 08 Posts: 7 |
Well, eventually I managed to do it. I updated the local configuration file - I relaxed the limits on how much disk space needs to be left free a little bit (and also freed some space) and also verified that the settings allow the machine to use two processors - one of these did the trick. Thanks to everyone who helped with ideas and suggestions! Micha. |
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.