Posts by Tuna Ertemalp
|
21)
Message boards :
BOINC Manager :
Logarithmic scale for the statistics graph?
(Message 62164)
Posted 14 May 2015 by  Tuna Ertemalp
Post: There is the "simple thing: better colouring for the statistics" thread here that deals with coloring the lines, choosing a palette, reassigning colors as people select/unselect projects, little popups on the lines as the user moves the cursor over the lines, and making the corresponding line more visible as the user moves the cursor over the project names. But, I did not see a suggestion about the scale of the graph; maybe I missed it. One problem I am seeing is that at the bottom of the graph, there are a lot of lines lying on top of each other. There are a few projects that hand out many credits, so they account for the lines that shoot into the sky, but then there are projects that are not too active, without many WUs, giving out less credit over time. Those get cramped at the bottom in a way that one ends up drawing on top of the other. Even selecting/unselecting a few doesn't make a visible difference. I have up to 53 projects on each of my machines (a few less if the CPU cannot run a VM, or the GPU is not strong enough for GPUGRID, or the Windows driver for the Intel Iris graphics turns my laptop into a brick while running Einstein) and the bottom of my graph is totally useless: http://1drv.ms/1PIwaaq It would be of great help if I had a chance to change the vertical scale to some sort of a logarithmic scale so that the lines bunched up at the bottom would have some room for separation, in addition to all the other suggestions in the other coloring thread (I especially like the one where the project line somehow becomes more visible as the cursor moves over project name on the right). Thanks for listening Tuna |
|
22)
Message boards :
BOINC Manager :
What is this: vbox_windows set to 0
(Message 62103)
Posted 10 May 2015 by Tuna Ertemalp
Post: And I've added it to Client configuration Thank you, sir! :) |
|
23)
Message boards :
BOINC Manager :
What is this: vbox_windows set to 0
(Message 62098)
Posted 9 May 2015 by Tuna Ertemalp
Post: I have it as the last entry in my cc_config.xml. I looked at the wiki, but couldn't find an explanation. Maybe I missed something... <vbox_window>0</vbox_window> |
|
24)
Message boards :
BOINC Manager :
Why does the BOINC Manager window move across boots from where I had left it?
(Message 62040)
Posted 3 May 2015 by Tuna Ertemalp
Post: To achieve a screen layout like that, my starting point (using Windows 7, I don't have 8.1 available to test yet) would be to open both the main Manager window and the Event log, then use Window's "Show windows side by side" tool (right click on task bar and select). That gives two windows of equal width, sharing the entire viewport between them. To obtain the unequal widths, I dragged the left side of the event log as far to the right as it would go, and dragged the right side of the main window to the right until it seemed visually aligned. Great suggestion. Never used that method before in so many years... :-) And Windows 7 doesn't have a Move function (correction: it does on 'Shift right-click'. How many people have found that?). There is also the ALT+Space to bring the Window System menu, then M to select Move even if you are not able to see it, and then you can use either the mouse or the cursor keys to relocate your window. Yeah, surely intuitive... :-) So, on balance, I think the BOINC developers have made a reasonably good stab at adopting a policy which is safe under multiple versions of Windows. Certainly. In the past I had BM windows lost outside the screen for whatever reason, and had to use my ALT+SPACE/M move. Still, I think allowing a window border width/height and, say, half a title bar height tolerance in the calculation would be nice. And perhaps the major fault is with Microsoft, for not aligning the hard edge when dragging a window border exactly with the edge-of-screen at XPos 0. Yup, agree. But, with all the shadows "outside" what a human would call a frame sort of blurs the idea of what exactly is the frame for the computer: The line we see, the outline of the drop shadow outside of that line that the computer needs to draw, what? Usability tests show preference for either. If you call the top left corner of the visible frame (0,0), then the corner of shadow/glow/etc. effects becomes negative coordinates. If you were to call that corner (0,0), then the "normal" frame would seem to be inset. With all the fancy drawing styles, you can't win. Hopefully Win10 will have the flat non-3D look, and (0,0) will be (0,0), period... Tuna |
|
25)
Message boards :
BOINC Manager :
Why does the BOINC Manager window move across boots from where I had left it?
(Message 62038)
Posted 3 May 2015 by Tuna Ertemalp
Post: So, I debugged it. Turns out there is a problem, but only shows up if you drag the left edge of your window all the way to the left, or right edge of the window all the way to the right. The location of the top and bottom edges, and the location of the upper left corner doesn't seem to matter. Looking at the registry, let's start with BOINC not running, and with these values for position (65,65), width (1800), height (800): [HKEY_CURRENT_USER\Software\Space Sciences Laboratory, U.C. Berkeley\BOINC Manager] "WindowIconized"=dword:00000000 "WindowMaximized"=dword:00000000 "Width"=dword:00000708 "Height"=dword:00000320 "XPos"=dword:00000041 "YPos"=dword:00000041 If you start BM, these values get respected and don't change when you exit BM. Good. Now, drag the top or bottom edge all the way to the top or bottom, respectively, which activates the "auto snap" in Windows, and BM snaps vertically to the top and bottom edges of my screen without a change in the horizontal. And exit BM. My registry shows 1800x1050 @ 65x0, which is correct: "WindowIconized"=dword:00000000 "WindowMaximized"=dword:00000000 "Width"=dword:00000708 "Height"=dword:0000041a "XPos"=dword:00000041 "YPos"=dword:00000000 Booting BM again puts the windows to that position, no problem. Now... Grab and drag the left edge of the BM window all the way to the left. Don't try to line it up with the edge of your screen or be slow and careful; just grab it and slam-drag it to the left until your mouse pointer can't move anymore, and let go of your mouse. That does not cause any "auto snap" by Windows (8.1 in my case), but (if you look closely) puts the left edge ever so slightly off the screen. Then exit BM, and registry shows something like: "WindowIconized"=dword:00000000 "WindowMaximized"=dword:00000000 "Width"=dword:00000750 "Height"=dword:0000041a "XPos"=dword:fffffff9 "YPos"=dword:00000000 Ah! Notice the xPos. It is trying to say "-7" but it can also be read as 4294967289, which would be clearly off the screen. In reality, this says the window is 1872x1050 @ -7x0 (1872, because 1800+65+7). Depending on how hard you slam the left edge of the window to the left side of the screen, the number at the end of fffffff might change, by the way. I got values between 9...D. Now, restart BM. You'll notice that the left edge is not off the screen, but instead there is a gap between the left edges of BM and the screen. Don't change anything and quit BM to get the values written to the registry: "WindowIconized"=dword:00000000 "WindowMaximized"=dword:00000000 "Width"=dword:00000750 "Height"=dword:0000041a "XPos"=dword:0000001e "YPos"=dword:00000000 This is 1872x1050 @ 30x0. In other words, BM didn't like 0xFFFFFFF9, and replaced it with a standard 30, shifting the whole window to the right by 37 pixels. Now, while BM is not running, change the registry back to the 1800x1050 @ 65x0 so that everything is within the screen, start BM, and now slam the right edge of the window to the right edge of the screen which puts the frame a few pixels off the screen. I am trying all this on a single monitor, by the way. Quit BM, look at the registry. Should be something like this (1858x1050 @ 65x0): "WindowIconized"=dword:00000000 "WindowMaximized"=dword:00000000 "Width"=dword:00000742 "Height"=dword:0000041a "XPos"=dword:00000041 "YPos"=dword:00000000 Now restart & quit BM, look at the registry again: "WindowIconized"=dword:00000000 "WindowMaximized"=dword:00000000 "Width"=dword:00000742 "Height"=dword:0000041a "XPos"=dword:0000003e "YPos"=dword:00000000 This is 1858x1050 @ 62x0. Notice how the window shifted left by 3 pixels since BM must have tried to correct for the right edge that was 3 pixels outside of the screen. So, that is the detailed description of the problem. I believe BM is making these little adjustments with good intentions, so that the BM window somehow doesn't end up outside the screen in a way that the user cannot grab it to move & resize. But, it should be more lenient to accommodate what I am running into by allowing a little bit to be left off the screen in all four directions. For that, BM should know what the window title bar height as well as window border height'n'width are, and allow part of them to be off the screen. The best way to find those sizes are, strangely, not to call GetSystemMetrics(), but instead use AdjustWindowRect(); you pass in some rect, and it returns to you the window size to accommodate that rect as the client window, and the difference between these two rects tells you everything you want to know about the width and height of title bar and window frame regardless whatever theme the user might be using: https://msdn.microsoft.com/en-us/library/windows/desktop/ms632665(v=vs.85).aspx By the way, BM stores the position & sizes of many other windows it has into registry, like the event log, and I am guessing they suffer from the same problem during restoring of these values. If the above gets fixed, it should get fixed for all windows. Thanks Tuna |
|
26)
Message boards :
BOINC Manager :
Why does the BOINC Manager window move across boots from where I had left it?
(Message 62031)
Posted 3 May 2015 by Tuna Ertemalp
Post: I nicely layout my BOINC Manager (snapped to left, almost maximized) and Event Log (snapped to right, with minimum width allowed, giving the rest of the screen to the main window) windows under Win 8.1 (but this has been happening in previous versions of BOINC Manager & Windows; I am just getting around to asking about it here). But, every time my machines get rebooted (or, I exit & restart BOINC Manager for whatever reason), the BOINC Manager window forgets where it was, and starts somewhere else. Example: After I lay them out: http://1drv.ms/1DUKjLU After I exit & restart the manager: http://1drv.ms/1KDVpK3 Windows apps should remember their exact size & position & [maximized | minimized | restored] state during quitting, and set them to the exact same values upon restart. Thanks Tuna |
|
27)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61695)
Posted 19 Apr 2015 by Tuna Ertemalp
Post: Well, the host who would ignore for over a week to roll up the most current Resource Sharing values it receives from the project servers, Update after Update, now has the correct RS values, all of a sudden. And, I didn't do anything on the host itself. I wasn't anywhere near it! So, mysteriously fixed... Still, I have the sched* and account* and client_state* files from before backed up if anyone needs information contained in them. Tuna |
|
28)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61659)
Posted 17 Apr 2015 by Tuna Ertemalp
Post: But, server code's age shouldn't matter, right? They have done their job and sent the right value to the host. At that point, it should be the host's job to interpret and roll up that value properly. All my other machines did that, but this host didn't and doesn't and won't. So, it must be a bug in BOINC Manager that happens only sometimes, and when it happens, it ignores the RS value coming from the server. Right? If I'm right, is there a BOINC developer listening in who would want me to turn on a few logging flags and do a few certain actions to generate a useful log before I try removing and re-adding these projects on this problematic host? |
|
29)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61654)
Posted 17 Apr 2015 by Tuna Ertemalp
Post: Also, if somebody knows the local XML file this is stored, maybe I can edit it manually? Soooooo.... The bottomline is, in all the sched_* and account_* XML files the Resource Share is 100 for all projects. So, projects do report stuff properly. But, both client_state and client_state_prev have the wrong values of 50...400. And, the timestamp of client_state is after the timestamps of sched_* and account_*. Also note the timestamps of the sched_* show that they are from the Update I had run earlier and reported in this thread. And, the account_* already had the right values from sched_* in the past, given the day early timestamp. But the client_* seems to not get them correctly from account_*. Thus, something seems to be going wrong during the processing of account_* into client_state. Nope, client_state is not locked or hidden or readonly or in use by some rogue process; I checked by renaming it to foobar.xml and then back to client_state.xml which would have failed if it were locked. 04/16/2015 11:49 AM 2,990 sched_reply_boinc.bakerlab.org_rosetta.xml 04/16/2015 11:49 AM 992 sched_reply_boinc.fzk.de_poem.xml 04/16/2015 11:49 AM 17,033 sched_reply_einstein.phys.uwm.edu.xml 04/16/2015 11:48 AM 1,103 sched_reply_lhcathomeclassic.cern.ch_sixtrack.xml 04/16/2015 11:49 AM 5,585 sched_reply_setiathome.berkeley.edu.xml 04/16/2015 11:48 AM 2,128 sched_reply_szdg.lpds.sztaki.hu_szdg.xml 04/16/2015 11:49 AM 4,335 sched_reply_www.enigmaathome.net.xml 04/16/2015 11:48 AM 1,113 sched_reply_www.malariacontrol.net.xml 04/16/2015 11:49 AM 5,134 sched_reply_www.primegrid.com.xml 04/16/2015 11:48 AM 18,544 sched_reply_www.worldcommunitygrid.org.xml 04/15/2015 09:35 PM 2,403 account_boinc.bakerlab.org_rosetta.xml 04/15/2015 09:33 PM 321 account_boinc.fzk.de_poem.xml 04/15/2015 09:36 PM 3,713 account_einstein.phys.uwm.edu.xml 04/15/2015 09:21 PM 340 account_lhcathomeclassic.cern.ch_sixtrack.xml 04/15/2015 09:36 PM 2,978 account_setiathome.berkeley.edu.xml 04/15/2015 09:33 PM 472 account_szdg.lpds.sztaki.hu_szdg.xml 04/15/2015 09:36 PM 1,014 account_www.enigmaathome.net.xml 04/15/2015 09:35 PM 585 account_www.malariacontrol.net.xml 04/15/2015 09:36 PM 2,933 account_www.primegrid.com.xml 04/15/2015 09:35 PM 2,440 account_www.worldcommunitygrid.org.xml 04/16/2015 02:53 PM 504,565 client_state.xml 04/16/2015 02:53 PM 509,753 client_state_prev.xml sched_reply_boinc.bakerlab.org_rosetta.xml:<resource_share>100</resource_share> sched_reply_boinc.fzk.de_poem.xml:<resource_share>100</resource_share> sched_reply_einstein.phys.uwm.edu.xml:<resource_share>100</resource_share> sched_reply_lhcathomeclassic.cern.ch_sixtrack.xml:<resource_share>100</resource_share> sched_reply_setiathome.berkeley.edu.xml:<resource_share>100</resource_share> sched_reply_szdg.lpds.sztaki.hu_szdg.xml:<resource_share>100</resource_share> sched_reply_www.enigmaathome.net.xml:<resource_share>100</resource_share> sched_reply_www.malariacontrol.net.xml:<resource_share>100</resource_share> sched_reply_www.primegrid.com.xml:<resource_share>100</resource_share> sched_reply_www.worldcommunitygrid.org.xml: <resource_share>100.0</resource_share> account_boinc.bakerlab.org_rosetta.xml:<resource_share>100</resource_share> account_boinc.fzk.de_poem.xml:<resource_share>100</resource_share> account_einstein.phys.uwm.edu.xml:<resource_share>100</resource_share> account_lhcathomeclassic.cern.ch_sixtrack.xml:<resource_share>100</resource_share> account_setiathome.berkeley.edu.xml:<resource_share>100</resource_share> account_szdg.lpds.sztaki.hu_szdg.xml:<resource_share>100</resource_share> account_www.enigmaathome.net.xml:<resource_share>100</resource_share> account_www.malariacontrol.net.xml:<resource_share>100</resource_share> account_www.primegrid.com.xml:<resource_share>100</resource_share> account_www.worldcommunitygrid.org.xml: <resource_share>100.0</resource_share> client_state.xml: <resource_share>50.000000</resource_share> client_state.xml: <resource_share>150.000000</resource_share> client_state.xml: <resource_share>300.000000</resource_share> client_state.xml: <resource_share>400.000000</resource_share> client_state.xml: <resource_share>300.000000</resource_share> client_state.xml: <resource_share>50.000000</resource_share> client_state.xml: <resource_share>200.000000</resource_share> client_state.xml: <resource_share>200.000000</resource_share> client_state.xml: <resource_share>300.000000</resource_share> client_state.xml: <resource_share>50.000000</resource_share> Then I got curious as to what other file might have ">400.000000<", focusing on World Community Grid. Boom! I got a hit in acct_mgr_request.xml which has this in it, along many more things, including the other wrong RS states. It is from about 25mins before me typing this post; it is set up such that my hosts contact BAM! every hour.
04/16/2015 10:43 PM 35,699 acct_mgr_request.xml
<project>
<url>http://www.worldcommunitygrid.org/</url>
<project_name>World Community Grid</project_name>
<suspended_via_gui>0</suspended_via_gui>
<account_key>blah blah blah</account_key>
<hostid>1363081</hostid>
<not_started_dur>0.000000</not_started_dur>
<in_progress_dur>0.000000</in_progress_dur>
<attached_via_acct_mgr>1</attached_via_acct_mgr>
<dont_request_more_work>0</dont_request_more_work>
<detach_when_done>0</detach_when_done>
<ended>0</ended>
<resource_share>400.000000</resource_share>
</project>
So, what is this file? Is this something BOINC Manager sends to the account manager? If so, it wouldn't be the culprit since it is simply telling the account manager what BOINC knows, without knowing what it knows is wrong, due to all the above stuff. Thus, a red herring. Or, is it? I don't really know what this file does... So, what else? Tuna |
|
30)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61653)
Posted 17 Apr 2015 by Tuna Ertemalp
Post: Never mind. Removed the files. I'll look at them. Seems I didn't fully read Richard's security warning in my haste to start my day. |
|
31)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61651)
Posted 16 Apr 2015 by Tuna Ertemalp
Post:
I thought I already sent this message, but I don't see it. Weird... Anyways, to get ready to investigate, I copied all relevant files to http://1drv.ms/1PUnbWv. I renamed files for projects that are misbehaving with a leading +, so they'll pop out better. I also included the result of the DIR of these files on my computer to get a snapshot of the timestamps. I'll investigate, yet I thought maybe someone else might want to see them instead of waiting for me to be done with my day. Tuna |
|
32)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61648)
Posted 16 Apr 2015 by Tuna Ertemalp
Post:
That's exactly what I like doing, anyways. So, I'll do the comparisons and post back. That'll probably take a little time though. So, later! Thanks for all the help so far, guys. I am curious. I'll be back. Tuna |
|
33)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61647)
Posted 16 Apr 2015 by Tuna Ertemalp
Post: Except for one host. That host got 100 for almost all of the projects, but a handful stayed at their old priorities. Stuck at: 50: Enigma PrimeGrid SZTAKI 150: LHC@home 200: Einstein SETI 300: MalariaControl Poem Rosetta 400: World Community Grid They are all up'n'about right now. Yet, an Update on each didn't do anything; see the log below. Mind you, this has been happening multiple times a day for the last week or so. So, the problem is not specific to today. BOINC latest: 7.4.42 Relevant log is below. Note that the fact that there are no tasks being requested shouldn't have any bearing on whether the Resource Sharing info is getting transmitted. Due to the large number of projects I run, all my machines are always in the state of not requesting tasks for more than 3-4 of my projects (seems part of the logic of how BOINC Manager rotates over the projects to satisfy the RS values and work buffers etc.), yet all their Resource Sharing values got updated the first time I issued a bulk Update on all the projects, where a good number of them didn't request tasks, yet got the "100". I guess I could also try to REMOVE these projects (except for PrimeGrid) via BAM! since they don't have any tasks in the queue, set PrimeGrid to NoNewTasks, update with BAM!, go find everything under the data folder with these project's names, delete them, then add them back via BAM!, hoping such a refresh will do the right thing. Then, I would repeat the same with PrimeGrid once it is drained of the tasks on this machine. Still, I am curious if we can figure out why this is happening. Might be an interesting bug somewhere in the software worth finding... 4/16/2015 11:48:31 AM | Enigma@Home | update requested by user 4/16/2015 11:48:31 AM | PrimeGrid | update requested by user 4/16/2015 11:48:31 AM | SZTAKI Desktop Grid | update requested by user 4/16/2015 11:48:31 AM | LHC@home 1.0 | update requested by user 4/16/2015 11:48:31 AM | Einstein@Home | update requested by user 4/16/2015 11:48:31 AM | SETI@home | update requested by user 4/16/2015 11:48:31 AM | malariacontrol.net | update requested by user 4/16/2015 11:48:31 AM | Poem@Home | update requested by user 4/16/2015 11:48:31 AM | rosetta@home | update requested by user 4/16/2015 11:48:31 AM | World Community Grid | update requested by user 4/16/2015 11:48:33 AM | SZTAKI Desktop Grid | Sending scheduler request: Requested by user. 4/16/2015 11:48:33 AM | SZTAKI Desktop Grid | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:48:36 AM | SZTAKI Desktop Grid | Scheduler request completed 4/16/2015 11:48:36 AM | | General prefs: from http://bam.boincstats.com/ (last modified 02-Apr-2015 22:18:55) 4/16/2015 11:48:36 AM | | Host location: none 4/16/2015 11:48:36 AM | | General prefs: using your defaults 4/16/2015 11:48:36 AM | | Reading preferences override file 4/16/2015 11:48:36 AM | | Preferences: 4/16/2015 11:48:36 AM | | max memory usage when active: 2455.82MB 4/16/2015 11:48:36 AM | | max memory usage when idle: 11051.17MB 4/16/2015 11:48:36 AM | | max disk usage: 30.00GB 4/16/2015 11:48:36 AM | | don't compute while active 4/16/2015 11:48:36 AM | | don't use GPU while active 4/16/2015 11:48:36 AM | | suspend work if non-BOINC CPU load exceeds 25% 4/16/2015 11:48:36 AM | | (to change preferences, visit a project web site or select Preferences in the Manager) 4/16/2015 11:48:41 AM | LHC@home 1.0 | Sending scheduler request: Requested by user. 4/16/2015 11:48:41 AM | LHC@home 1.0 | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:48:43 AM | LHC@home 1.0 | Scheduler request completed 4/16/2015 11:48:48 AM | malariacontrol.net | Sending scheduler request: Requested by user. 4/16/2015 11:48:48 AM | malariacontrol.net | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:48:50 AM | malariacontrol.net | Scheduler request completed 4/16/2015 11:48:55 AM | World Community Grid | Sending scheduler request: Requested by user. 4/16/2015 11:48:55 AM | World Community Grid | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:48:56 AM | World Community Grid | Scheduler request completed 4/16/2015 11:49:01 AM | rosetta@home | Sending scheduler request: Requested by user. 4/16/2015 11:49:01 AM | rosetta@home | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:49:02 AM | rosetta@home | Scheduler request completed 4/16/2015 11:49:08 AM | Enigma@Home | Sending scheduler request: Requested by user. 4/16/2015 11:49:08 AM | Enigma@Home | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:49:10 AM | Enigma@Home | Scheduler request completed 4/16/2015 11:49:15 AM | SETI@home | Sending scheduler request: Requested by user. 4/16/2015 11:49:15 AM | SETI@home | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:49:16 AM | SETI@home | Scheduler request completed 4/16/2015 11:49:21 AM | Einstein@Home | Sending scheduler request: Requested by user. 4/16/2015 11:49:21 AM | Einstein@Home | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:49:23 AM | Einstein@Home | Scheduler request completed 4/16/2015 11:49:44 AM | Poem@Home | Sending scheduler request: Requested by user. 4/16/2015 11:49:44 AM | Poem@Home | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:49:46 AM | Poem@Home | Scheduler request completed 4/16/2015 11:49:51 AM | PrimeGrid | Sending scheduler request: Requested by user. 4/16/2015 11:49:51 AM | PrimeGrid | Not requesting tasks: don't need (CPU: not highest priority project; NVIDIA GPU: job cache full) 4/16/2015 11:49:54 AM | PrimeGrid | Scheduler request completed |
|
34)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61645)
Posted 16 Apr 2015 by Tuna Ertemalp
Post: Sorry for thinking Priority equals Resource Share. But that is what the grid column says in BOINC Manager... :-) Yeah, about that... It was 3am, I was barely awake, on my smartphone, typing away, away from the PC, and I had this strong image of "Pri..." as the truncated form of column title "Priority" in my head. No idea where that brain fart came from. I was soooo sure... |
|
35)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61640)
Posted 16 Apr 2015 by Tuna Ertemalp
Post: I'd guess you'd have to update every project before the RS transfer is complete Already done, as stated before. I updated every project on all of my machines; everything worked on all my machines, except for this one where only a subset of the projects are not getting the new Resource Share value. The preferences aren't updated directly from BAM. BAM updates the preferences to the projects and then the client updates the preferences from a project. In order to update the preferences of the client after you have edited the preferences in BAM, press the "Save to projects" button. Then update a project (no matter which) and you will receive the updated preferences. That already worked, as stated before. Every project knows it is at 100. I verified that further by spending the time to visit every project website. Plus, since my other 7 machines got the right "100" during their Update process of each project, obviously they are correct at the project websites, thus the values did travel from BAM! to project websites properly. But, doublechecked. So, the end result with the problematic projects is that the project web sites display one value (100) for resource share, and BOINC Manager still displays a different value, even after you do an 'Update' so that the client contacts the project and gets any updated information? Yes. I actually should have trimmed my story to the bare essentials, not overshare :-), and say: "All my projects used to have different Resource Sharing values but now they all have a setting of 100, correctly displayed at project websites. All of my hosts got this 100 for each project during their project updates, except for one host that is stuck with the old values for a subset of projects even after repeated project updates." Taking only the problematic projects: do you have any other 'Venues' (groups of preference settings - default, home, school, work - aka 'host locations') set up? If so, is the RS set at 100 for all of them? Can you match the number shown in BOINC Manager with anywhere else? If you call up the 'Properties' page for the project in BOINC Manager, does the 'Host location' show what you expect? (Note that you can change the host location project-by-project, but only via the project website or possibly BAM! - there's no local override) No other Venues. Can't match the old values to anything else on the project website under my account. All of my Host Locations are shown as "default" in the BOINC Manager and "---" on the project websites. And when I check the Machines page on each project website under my account, Location column is blank. So, blank = default = '---'. So, somehow this machine is not respecting the values it is getting for these projects from the project sites during the project update process. Or, the projects are sending the "old" value to this host only, but that "old" value is nowhere on their project websites. Question is why. Also, if somebody knows the local XML file this is stored, maybe I can edit it manually? The only other thing I can think of is to uninstall BOINC, remove all the data, and reinstall. But, I have so many really long PrimeGrid jobs that are finished 90%, yet will take weeks to finish the next 10%. And, PrimeGrid is one of those projects on this machine that is stuck... :-( I would hate to lose all that work... |
|
36)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61628)
Posted 16 Apr 2015 by Tuna Ertemalp
Post: I think anything you set to 100 via BAM! would be called a "Resource Share" in BOINC terminology. Priorites (especially in the context of Event Log debug flags) are something different entirely. Hmmm... My understanding of how this works is: BAM! tells each project PRJ "Hey PRJ, your Resource Share for user Foo is now NNN", and its job is done. Then the BOINC Manager on each host HST maintained by user Foo eventually goes "Hey, I'm running an Update on PRJ; so PRJ, what is the Resource Share you've got for my master Foo?" Then PRJ responds with NNN, BOINC Manager says thank you, and uses that NNB as PRJ's Resource Share. In my case, BAM! told 44 projects that their Resource Share for me is 100. They heard and accepted it; I can confirm it on the project sites under Project Preferences under Your Account. And, 7 of my 8 hosts got the 100 during their very next Update of PRJ. The 8th host heard the news for about 35 projects, and insists on turning a deaf ear to 8 or 9 (I'm away from my computers right now, thus inexact count). Am I wrong? If not, then the issue is with my local BOINC software/data on this host that ignores what some of my projects return. Tell me how to find out what is broken, and how to fix it. If I'm wrong, tell me how it really works. Sorry for thinking Priority equals Resource Share. But that is what the grid column says in BOINC Manager... :-) Cheers... Tuna |
|
37)
Message boards :
Questions and problems :
Why does one of my hosts have some projects' priorities stuck at their past values
(Message 61625)
Posted 16 Apr 2015 by Tuna Ertemalp
Post: I had assigned different priorities to different projects for quite some time. Then, recently, I said "oh, whatever", and set them all to 100 using BAM!. All my machines use BAM! as the account manager. I forced an update of all the projects on each host, and they all got 100s for the projects they churn. Great. Except for one host. That host got 100 for almost all of the projects, but a handful stayed at their old priorities. No amount of updating helps. Since those projects currently didn't have any tasks churning on that host, I also did a reset and then update, nada. Still the old values stuck around. I turned on the priority_debug for the Event Log, but to my eyes, uninitiated in the dark magic of BOINC's inner workings, I couldn't understand if it gave any clues. Clearly, projects themselves know what their new priority for me is, i.e. BAM! has told them successfully that whatever they had for me changed to 100. What would a more knowledgeable person want to figure this out? Turn on another flag? How much of the tail portion of stdoutdae.txt would be useful after taking which actions? Thanks Tuna |
|
38)
Message boards :
Questions and problems :
How does BOINC decide to ask for tasks???
(Message 61553)
Posted 12 Apr 2015 by Tuna Ertemalp
Post: Thanks! |
|
39)
Message boards :
Questions and problems :
How does BOINC decide to ask for tasks???
(Message 61547)
Posted 12 Apr 2015 by Tuna Ertemalp
Post: Obviously, I will at least turn on the one you mentioned. If anybody knows any other useful flag, speak now... :-) |
|
40)
Message boards :
Questions and problems :
How does BOINC decide to ask for tasks???
(Message 61546)
Posted 12 Apr 2015 by Tuna Ertemalp
Post: Cool. I have 7 other Windows machines currently draining, with the same set of projects, same 100 weight each, all at NoNewTasks state. What log flags would you like me to turn on to get useful stuff in the logs before I do my next SelectAll/AllowNewTasks, possibly as early as tonight or tomorrow (GMT-8)? |
Previous 20 · Next 20
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.