 |
|
|
1)
Message boards :
Number crunching :
post crash duplicate credit
(Message 23200)
Posted 18 Jun 2006 by old_user9685 Post: I\'m currently running 2 models on a HT box and they both crashed on 13th June. The last backup I had was 3rd June, so I restored to that, removed all the pending trickles and zips, edited the xml to indicate that they\'d been uploaded (they had been on 10th, 11th & 12th June) and wrote off 10 days worth of crunch time. Nothing unusual so far. Since I didn\'t want to miss a trickle upload, I decided to just allow BOINC to upload any new trickles generated during the 10 day recovery period. The server is supposed to just ignore them and it is, since the trickle data for the results remains static at 12 June. Problem is that the stats sites show my credit has increased at around the same time as these \"duplicate\" trickles are being uploaded. This is wrong. I guess I could try and stop it by removing the trickles, but this whole issue is contra to my understanding of what\'s supposed to be happening, so I\'m a little nervous of doing that now and if there\'s a hole here, it needs plugging. Help? resultid=5360066 resultid=5324594 |
|
2)
Questions and Answers :
Windows :
Downloaded program and nothing happens?
(Message 18094)
Posted 12 Dec 2005 by old_user9685 Post: I did wait for the entire time. The files downloaded up to the 100% and then they disappeared and never appeared on the work tab Don, which version of BOINC are you running? If you\'re using v4.45 of BOINC, you\'d be better advised to drop it for a later version. The v4.45 has a rather annoying download flaw that could be tripping you up. |
|
3)
Questions and Answers :
Windows :
BOINC 5.2.5 released
(Message 17109)
Posted 10 Nov 2005 by old_user9685 Post: Did that behaviour start with 5.2.7 or any 5.2.x version? Only tried 5.2.7 on this particular system.
On other systems both 5.2.1 & 5.2.2 also suspend network. [edit] But they are \"always on\" & have no dial-up capability configured. [/edit] It\'s probably something specific to my setup, so here it is if anyone cares to replicate: System is XP pro, SP2 Dial-up networking configured with 4 dial-up entries. 10/100 Network card installed, media disconnected. System running BOINC 4.45 (CPDN & Rosetta) installed as service with network disabled until some trickles are pending. Starting with 4.45 running. Suspend BOINC. Stop Service. Uninstall 4.45 by launching boinc4.45 installer and selecting \"remove\". Install 5.2.7 as service. Observed: BOINC tries to trickle up, even before completing the mandatory version change benchmark. BOINC continues trying to trickle up even though network access is suspended. Knock yourselves out. ;-) |
|
4)
Questions and Answers :
Windows :
BOINC 5.2.5 released
(Message 17078)
Posted 9 Nov 2005 by old_user9685 Post: And we have a new version 5.2.7 available from the same download location I gave it a whirl at home last night, and I\'m not impressed. I use a modem for connectivity there and even with n/w access suspended, it continually tries to send trickles and can\'t resolve the cpdn hostname. Also with n/w access suspended, it was popping up a box asking if \"it\'s OK to connect\". The network options weren\'t being remembered. Regardless of what I set, it just went back to \"Automatically detect\". Then after I\'d connected and allowed network connectivity, the daemon (installed as a service) just died with no explanation in the middle of my uploads and left the manager connectionless. I remember the days when n/w access disabled meant exactly that. It included trickles, master file fetches, scheduler requests and any other client/server communication, not just uploads and downloads. Disabled means DISABLED damnit! :( Needless to say, I went back to 4.45 ! Perhaps 5.2.7 may be the recommended version for \"always on\" systems, but modem users beware. The irony is of course that, if I remember correctly, not long ago the v5.x series was touted to be the modem users\' best friend. This may be true of the scheduler code, but the networking code is killing it for me. [rant mode off] I\'ve long been a staunch supporter of the BOINC concept, but if 5.2.7 is anything to go by, it is really starting to look like we\'re heading 180 degrees off course. It\'s just too buggy for my liking. I can\'t even defend it as \"alpha\" or \"beta\" software because it\'s the recommended version! At this point I can only feel sorry for new users. |
|
5)
Questions and Answers :
Windows :
BOINC 5.2.5 released
(Message 17032)
Posted 7 Nov 2005 by old_user9685 Post: If your prefs are set to remove from memory, then the benchmark removes the app and you still have the 50/50 chance of an abort. I haven\'t tested extensively tbh. I did some testing with the public project and the results echo your own tests. Benchmarks abort, but BOINC now appears to catch the \'orphaned\' app and restart it. I would like to do more testing, especially with the sulphur models since they seemed to be more susceptible to this problem. As soon as I get the time...:) |
|
6)
Questions and Answers :
Windows :
BOINC 5.2.5 released
(Message 16919)
Posted 1 Nov 2005 by old_user9685 Post: It is supposed to be. The same delay as is in the 4.45b version is supposed to be standard now. No. No. The delay is the same, still 10secs. The difference is that the 5.x version now honours your preferences regarding \"leaving apps in memory\' at benchmark time. If your prefs are set to leave in memory, then the benchmark does not remove the app (and can\'t trip over the 10sec timeout). If your prefs are set to remove from memory, then the benchmark removes the app and you still have the 50/50 chance of an abort. Moral of the oral: When switching from 4.45b to a v5.2.x client, also make sure that you change your preferences to \"Leave applications in memory\". (And don\'t forget to click update from the client to get the updated preferences....) |
|
7)
Questions and Answers :
Windows :
Dialup or Broadband connection
(Message 16918)
Posted 1 Nov 2005 by old_user9685 Post: Hi Peter You can try going to the connections tab on the options dialog. Once there you can try \'auto\', \'lan\' or \'dial-up\'. Perhaps LAN is the one for you? Sorry, not much knowledge of broadband. |
|
8)
Message boards :
Number crunching :
Help - Am In A Loop
(Message 16673)
Posted 19 Oct 2005 by old_user9685 Post: I cannot download all of that in the time allowed on my dial-up Wanadoo. So I had to disallow internet connection after a while. On resuming, the sulphur files continued to dowmload where they had left off, but the computer reported in red that the work unit had failed with a download error, and I must wait for another day to get a new one. v4.45 contains a bug in the networking code. The wu file itself is quite small, but the associated files are large (exe\'s, tga\'s, zip\'s). If a download of any of the associated files is interrupted (disable n/w access, modem line drop, computer freeze etc.) the wu immediately errors out. In order to get a wu running successfully, all of the files have to download uninterrupted. This problem frustrated me for a long time as well, especially since I had very unreliable modem connectivity. You have basically two options. a.) Revert to an earlier version or to v5.2.x, as already mentioned. b.) Implement the v4.45b patch available here BOINC 4.45b. This version is mostly known for it\'s extended benchmark timeout fix, however it also contains the fix to the network problem described above. Using either of these two options you will be able to extend the 31MB download over multiple sessions. I recommend the following approach to the upgrade. Start BOINC when disconnected. [1] Disable BOINC network access. [1] Reset the CPDN project. [2] Exit BOINC. [3] Upgrade / downgrade to a fixed version. Start BOINC. Establish your internet connection. Enable BOINC network access. [4] [1] You do not want to start downloading anything until the appropriate time. [2] This clears any pending incomplete associated file downloads. If you do not reset the project, any leftover large files will still be downloaded and if you don\'t have a successful wu ready to crunch, these large files will be deleted and you will have to re-download them again in 24hours time. Save yourself some wasted bandwidth. [3] Stop the service if BOINC is running as a service. [4] If you have reached your quota of CPDN wu\'s, a message to that effect will be displayed and nothing will be downloaded. You can then reconnect after the necessary waiting interval. If anything isn\'t clear, please post a follow-up. Good Luck! |
|
9)
Message boards :
Number crunching :
Computational Error
(Message 16672)
Posted 19 Oct 2005 by old_user9685 Post: Regarding preferences: On your preferences settings, set \"Leave applications in memory while preempted?\" to yes. This will prevent the model from being unloaded each time BOINC switches between projects, and you therefore have less chance of a -5 error when the model \"restarts\" because it will resume instead of restart. Regarding BOINC clients: You have two choices for clients. v5.2.x of BOINC will not unload the model from memory when benchmarking (unless \"Leave applications in memory while preempted?\" is no), so using this client is peferable to avoid the problem described in your initial post. v4.45 of BOINC will unload the model when benchmarking, but only waits 10 secs for the model to terminate. Should the model take more than 10 secs to terminate, BOINC will abort the benchmarks and you\'re likely to have your system running idle until you notice it. The \"unofficial\" v4.45b that Les has indicated waits 30 secs (avoiding the idle state), but the model is still unloaded and you still run the risk of it dying on restart. If you\'re determined to stay with a v4 client, then this is the suggested one for the reasons I have explained above. |
|
10)
Questions and Answers :
Windows :
BOINC 4.45 incompatibility with Backup4all
(Message 16599)
Posted 14 Oct 2005 by old_user9685 Post: The boinc.exe daemon process in v4.45 now listens on port 1043 for boinc client connections. Perhaps the client or server part of your backup application also uses this port? If this is the case, the backup application developers ought to consult the IANA well known port list, which shows the registered boinc ports and they would be well advised to register their own ports to prevent future similar situations. |
|
11)
Questions and Answers :
Windows :
message: Invalid account file: account_.xml
(Message 16071)
Posted 16 Sep 2005 by old_user9685 Post: I suspect that this is because you have a file or folder in your BOINC folder that starts with \'acc\', but is not a BOINC account file. NOTE: You must have some XML files that are named \'account_[URL OF PROJECT].xml\' where [URL OF PROJECT] is the project url. e.g. the one for climateprediction is: \'account_climateprediction.net.xml\' If you have any other files or folders that start with \'acc\', but are not of the type listed above, then try stopping BOINC, renaming the file/folder and starting BOINC again. (Ref: CVS, David 27 June 2005, account folders) If you actually have a file named \'account_.xml\' in your BOINC folder, stop BOINC rename it to \'tmp_account_.xml\' and restart BOINC. After you\'ve confirmed that everything else is normal, then delete that file. (Ref: CVS, David 27 Feb 2005). Telling us your BOINC version may help also... ;-) |
|
12)
Questions and Answers :
Windows :
Benchmak stopped model being crunched
(Message 15273)
Posted 21 Aug 2005 by old_user9685 Post: > There is a problem with the benchmarks run by this version. They are very low. There's a new version (4.45b) on its way to Arnaud. Benchmark issue appears sorted. Holler if you find otherwise. |
|
13)
Questions and Answers :
Windows :
Benchmak stopped model being crunched
(Message 15219)
Posted 19 Aug 2005 by old_user9685 Post: > To get 4.45a with a 30 sec delay instead of 10, There is a problem with the benchmarks run by this version. They are very low. I haven't yet figured out the problem (hoping just missing optimizations), so please don't download this version. If you already have, I would suggest reverting back to the UCB 4.45 until I have had a chance to figure out the problem. Sorry all. Chris :( |
|
14)
Questions and Answers :
Windows :
Benchmak stopped model being crunched
(Message 15163)
Posted 17 Aug 2005 by old_user9685 Post: > The timeout for stopping applications for the benchmark run has been > increased, hopefully enough to allow CPDN to stop in time. What is apparently In the latest dev branch (4.72), the timeout is still at 10 seconds. Unless someone is still planning to increase it, it's not likely to be in the next release. |
|
15)
Message boards :
Number crunching :
Server/Trickle Problems Being Worked On
(Message 13996)
Posted 1 Jul 2005 by old_user9685 Post: > Yep, I have two (one finished on the 20th, and another on the 21st) that > completed with 72 trickles, but the end of phase 3 graph is not there, and the > credits are incorrect for a successful run. > > http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=870517 > http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=856478 ditto. Just in case it's worth mentioning: Finished on 20th June, 72 trickles, Outcome Success, No P3 Run Info, odd credit http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=862571 I also have one where everything looks okay, but server state still in progress http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=769760 I presume that if there are any problems with these models server-side, someone will contact me via email... :) |
|
16)
Message boards :
Number crunching :
Boinc V4.25 Stable Windows Version Released
(Message 11166)
Posted 20 Mar 2005 by old_user9685 Post: > into the service and set it to run as a local system account (why the > installer doesn't allow this option I'm not sure). I've been reading about It doesn't use LocalSystem by design, as a safety precaution. If it were to use that account, then BOINC and the underlying science app would have elevated privleges on your system, which as a default is no go. Nothing stopping anyone from changing it after installation though, but if something goes wrong (i.e. you get hacked by a dodgy project application), the finger of blame points away from the system designers. That's the theory anyways. |
|
17)
Questions and Answers :
Windows :
4.19 \"No schedulers responded\" behind http proxy
(Message 9901)
Posted 24 Feb 2005 by old_user9685 Post: > You're going to have big problems if you upgrade to a development version as > they require a signature on all the downloaded files. CPDN doesn't have a > signature on the 3 hadsm3*.zip files, and BOINC will reject them. This will > continue to be the case until the server side software is rebuilt using the > current BOINC development source. Thanks Thyme Lawn You are right about the downloads. If you're busy with a model and need to trickle up, the development version could work for you. If you're nearing the end of a model, you should probably let the model complete and suspend network activity until the CPDN server software has been updated. Downloading a new model (or work from most BOINC projects atm.) will fail, as explained in the thread noted by Thyme. Hope this doesn't confuse everyone too much. :-/ |
|
18)
Questions and Answers :
Windows :
4.19 \"No schedulers responded\" behind http proxy
(Message 9894)
Posted 24 Feb 2005 by old_user9685 Post: Hi Lewelma > Any news on this issue? I am having the exact same problem as outlined below. Apologies for the delay. Yes, the problem has been found and corrected in the current BOINC client development branch. The BOINC team suggest that if you are suffering from proxy related problems that you use the later version of the software: [Quote from <a href="http://boinc.berkeley.edu/download.php">Download BOINC client software</a>] Version 4.19 (released 25 Jan 2005) This version doesn't work with some HTTP proxies. If you use a proxy and experience problems, please use <a href="http://boinc.berkeley.edu/download.php?dev=1#dev">version 4.23</a>, which fixes this problem. [End Quote/] Actually v4.23 has been replaced with v4.24 at the moment. > I've also heard that the Oxford server had some issues but has since been > repaired. However, this does not seem to have fixed the "No schedulers > responded" issue (at least for me!). Yes, that's fixed from v4.23 onwards. :) Please remember that these later versions are a bit different from what you may be used to, and being Alpha software, they could possibly introduce other problems. In any event, it looks like the new versions may be released to the public quite soon. Fingers crossed. |
|
19)
Questions and Answers :
Windows :
4.19 \"No schedulers responded\" behind http proxy
(Message 9691)
Posted 21 Feb 2005 by old_user9685 Post: > Please could you email me again with your email address! done. |
|
20)
Questions and Answers :
Windows :
4.19 \"No schedulers responded\" behind http proxy
(Message 9474)
Posted 17 Feb 2005 by old_user9685 Post: > Which data do you want to see, from which log file? Basically all the [DEBUG_HTTP ] data in the stdout.txt file from 2005-02-17 12:44:23 [climateprediction.net] Sending request to scheduler: http://climateapps2.oucs.ox.ac.uk/cpdnboinc_cgi/cgi until 2005-02-17 12:44:27 [climateprediction.net] Deferring communication with project for 1 minutes and 0 seconds I'll email you privately and request that you send me the data, rather than post it all here. Thanks |
©2024 climateprediction.net