Questions and Answers :
Windows :
Decommissioning a machine
Message board moderation
Author | Message |
---|---|
Send message Joined: 7 Sep 04 Posts: 14 Credit: 160,054 RAC: 0 |
I\'ve got a work unit approaching phase 3 completion on a machine which I will shortly be unable to use further for BOINC. I\'d like to ensure that the model finishes and uploads the results, but I don\'t want BOINC to go ahead and download a new model on that machine. What\'s the best way to achieve this? |
Send message Joined: 5 Aug 04 Posts: 1283 Credit: 15,824,334 RAC: 0 |
The best way to do it would be to reduce <b>Use no more than</b> in your general preferences so that the disk space available to BOINC is only around 500MB more than is currently used for your BOINC projects. |
Send message Joined: 7 Sep 04 Posts: 14 Credit: 160,054 RAC: 0 |
> The best way to do it would be to reduce <b>Use no more than</b> in your > general preferences so that the disk space available to BOINC is only around > 500MB more than is currently used for your BOINC projects. > Thanks for the suggestion. That sounds like it might work, but won't the upload of the previous model free up as much space (or maybe even more space) than that required for the download of a new model? So unless I reduced the space to less than I'm currently using, I'd have thought it might still be able to take the unit. I don't know what would happen if I throttle disk usage in my general preferences to a level <b>below</b> my current usage... (and I guess I'd also have to consider the impact on other BOINC machines, making sure I set the preference back again before any of them polled the server and picked up the changed preference). One advantage I have with the machine in question is that it is mostly off line so it isn't going to start any uploading until I connect it to the network. This gives me the opportunity to control the exact time of the upload which should help. |
Send message Joined: 5 Aug 04 Posts: 1283 Credit: 15,824,334 RAC: 0 |
Upload doesn't free up disk space as CPDN keeps the results on your system, but they are compressed down to about 330Mb. On reflection, setting your disk space to 500Mb more than you're currently using might be a bit too generous. A better idea would be to subtract the disk space used for the system's current model from the disk space used by BOINC and add 800Mb. That'll be enough to complete the result, compress it and prevent another model from being downloaded. I would suggest putting the machine you want to decommission into another venue so it doesn't affect any other systems, but there's a bug in BOINC that prevents venue changes from being propogated down to the clients (but it is possible to manually edit the venue into client_state.xml when BOINC isn't running). What you could do is delay changing the disk limit until you're closer to completing the model, disable network access for all your other systems, change your disk space preferences and update preferences on the system you're decommissioning. <br><a href="http://www.teampicard.net"><img src="http://www.teampicard.net/templates/fisubice/images/phpbb2_logo.jpg"></a><a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/team_display.php?teamid=3">Join us here</a> |
Send message Joined: 2 Sep 04 Posts: 44 Credit: 372,682 RAC: 0 |
I find that using <b>Leave at least</b> instead of <b>Use no more than</b> is an easier option. Set it to a ridiculous figure (500GB) and when BOINC requests more work (which can occur a few days before your current model is uploaded, depending on your <b>Connect to network about every</b> setting) you will see a message along the lines of "Message from server: No work available (there was work but you don't have enough disk space allocated)" I have never seen the <b>Leave at least</b> setting have any impact on running projects. It only affects the client when it tries to obtain new work. Your current model should be uploaded without any problem. (If they can get that Swiss server running again, but that's another story...) Make sure that you update your CPDN project from the client so that the new preferences are applied, before it requests more work. I would suggest the following approach: a.) Change your general prefs <b>Leave at least</b> setting to a value higher than the disk space on your system. b.) As soon as the update has been confirmed on the server, perform a project update from the client in question. When you see the message: General prefs: from CPDN (last modified YOURDATE YOURTIME) in your messages tab, disable network access on that client. c.) Change your general prefs <b>Leave at least</b> setting back to your normal setting. d.) Wait until the climate model has completed. e.) Go back to step a.) but enable network access in step b.) instead of disabling it. Once you've enabled the network access the model will upload and BOINC will not request any more work from CPDN. Once the upload has completed successfully, detach from CPDN. This assumes that you're only running one project on that host. If not, you could follow these steps for each project, detaching from each one once you've completed their uploads. If you are running any other projects, you can experiment by changing the settings on another project, since they all have short running WU's. <i>Just don't forget to set the <b>Leave at least</b> setting back to normal after you've updated the client, otherwise your other projects/hosts will be affected.</i> Kind Regards |
Send message Joined: 7 Sep 04 Posts: 14 Credit: 160,054 RAC: 0 |
Thanks for all your help folks. I'm trying the "Leave At Least" technique. I believe I've now successfully convinced the machine in question to leave 500 GB, taken it back off line and restored my preferences to 'proper' values for my other BOINC machine. Now I just have to wait for the model finish... I wonder why BOINC holds these preferences centrally? Seems to me that the resource limits a user needs to set, more properly belong at the level of an individual machine and would be better set by the client and read by the server, rather than the other way around. If preferences could be set on a machine by machine basis (or maybe overridden on a machine by machine basis if you wanted a centralised "default" for users of many machines) you'd have a more flexible system, better able to cope with issues like this one. Just some idle design thoughts.... |
Send message Joined: 2 Sep 04 Posts: 44 Credit: 372,682 RAC: 0 |
> Thanks for all your help folks. I'm trying the "Leave At Least" technique. I > believe I've now successfully convinced the machine in question to leave 500 You can always check this by going to the disk tab and reviewing your 'BOINC free space' setting. It should be 0 while trying this technique. > GB, taken it back off line and restored my preferences to 'proper' values for > my other BOINC machine. Now I just have to wait for the model finish... Please don't forget to set it to 500GB <i>before</i> you enable network access on the machine in question, otherwise during the course of the upload (which can take a bit of time) it may send a scheduler request, get your non-restrictive settings and start downloading another CPDN wu. Good Luck. |
Send message Joined: 21 Sep 04 Posts: 4 Credit: 13,299 RAC: 0 |
David Worton seems to have succeeded.. (he has very nice UL/DL rates) What about this WU? http://climateapps2.oucs.ox.ac.uk/cpdnboinc/workunit.php?wuid=25560 Exit status : -187 (0xffffff45) stderr out : 4.13 Output file exceeded size limit The 72 trickles and the graphs are there, I'm guessing this is the archived results? Would it be caused by the Leave at Least setting? <img src=\"http://boinc.mundayweb.com/one/stats.php?userID=6\"> |
Send message Joined: 7 Aug 04 Posts: 2184 Credit: 64,822,615 RAC: 5,275 |
> What about this WU? > > http://climateapps2.oucs.ox.ac.uk/cpdnboinc/workunit.php?wuid=25560 > > Exit status : -187 (0xffffff45) > stderr out : 4.13 > Output file exceeded size limit > > The 72 trickles and the graphs are there, I'm guessing this is the archived > results? Would it be caused by the Leave at Least setting? > See the most recent post (and links in that post) in this thread: http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=1115 for more information. |
©2024 cpdn.org