| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: WCG going beyond its resource restrictions/paralyzing machine |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 10
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi! I've been using WCG on a few different machines over the years, and all has been well til I got a Win 7 Pro (64 bit) machine from work. WCG has not agreed with this machine at all.
----------------------------------------Despite setting WCG to use no more than 45% of CPU time (I have dialed this down further, but it seems to make no difference), if WCG is running along with Thunderbird or Chrome with a lot of windows open, or both, my CPU will periodically flatline at 100% and everything will freeze up for long periods of time, with the WCG processes definitely adding up to well over 45%. It seems to coexist particularly poorly with Thunderbird. Generally, pausing WCG seems insufficient to remedy the issue, and boinc.exe seems to keeps hitting the CPU in that case. If I close WCG entirely, which I can accomplish only haltingly, things will go back to normal in about 10 seconds. I should also add that, about 10% of the time, when I close WCG under these circumstances, the problem does NOT subside--and at that time, I can look in the Task Manager and see that there is a rogue work unit (or even two!) hitting the CPU even after WCG is closed! I have to close that work unit manually using Task Manager. Not sure how this is possible. After I noticed this behavior, I went from 4 to 8 GB of RAM, but that has not noticeably changed things. Any suggestions? My WCG crunching time has drastically decreased, given that I can really only crunch without much else running on the machine. Thanks muchly! [Edit 1 times, last edit by Former Member at May 12, 2012 12:28:40 AM] |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello sevenveils,
This is unlike any reports from other people, so I will have to ask you for a lot of information. But first, let me explain that CPU usage of 45% will be truncated to 40%, which will mean running 100% for 2 seconds out of 5, then running 0% for 3 seconds out of 5. BOINC is a multi-OS program, so it does not include OS-specific code which would be needed for fine control. The first thing is to post the first page of your BOINC Messages tab after a startup. That will tell us a lot about your system. Next, investigate the rogue WCG projects. Bring up Task Manager, move it to one side, then page down to the WCG projects. Then bring up BOINC Manager and go to the Projects tab. Select WCG, then suspend. Do all the WCG projects stop in Task Manager? If so, try resume. If trouble never shows -- well, try and make this problem show up somehow. I have no idea how it can happen. This problem is so surprising, I hope you post the Messages tab before you try to locate this problem.  Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi--thanks for writing. I should edit that the process that remains on pause (which requires that I actually shut down WCG) is boincmgr.exe, not boinc.exe.
----------------------------------------As far as the rogue work units, that generally happens when I am quitting WCG under this very paralyzed condition. It takes a long time to close, and seems that sometimes it doesn't manage it completely. Here's the Messages startup text, thanks for your help! 5/11/2012 8:23:57 PM Starting BOINC client version 6.10.58 for windows_x86_64 5/11/2012 8:23:57 PM log flags: file_xfer, sched_ops, task 5/11/2012 8:23:57 PM Libraries: libcurl/7.19.7 OpenSSL/0.9.8l zlib/1.2.3 5/11/2012 8:23:57 PM Data directory: C:\ProgramData\BOINC 5/11/2012 8:23:57 PM Running under account XXXX 5/11/2012 8:23:57 PM Processor: 4 GenuineIntel Intel(R) Core(TM) i5 CPU M 560 @ 2.67GHz [Family 6 Model 37 Stepping 5] 5/11/2012 8:23:57 PM Processor: 256.00 KB cache 5/11/2012 8:23:57 PM Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 nx lm vmx smx tm2 popcnt aes pbe 5/11/2012 8:23:57 PM OS: Microsoft Windows 7: x64 Edition, Service Pack 1, (06.01.7601.00) 5/11/2012 8:23:57 PM Memory: 7.80 GB physical, 15.60 GB virtual 5/11/2012 8:23:57 PM Disk: 232.88 GB total, 65.48 GB free 5/11/2012 8:23:57 PM Local time is UTC -7 hours 5/11/2012 8:23:57 PM No usable GPUs found 5/11/2012 8:23:58 PM World Community Grid URL http://www.worldcommunitygrid.org/; Computer ID 1765920; resource share 100 5/11/2012 8:23:58 PM World Community Grid General prefs: from World Community Grid (last modified 07-May-2011 13:19:19) 5/11/2012 8:23:58 PM World Community Grid Host location: none 5/11/2012 8:23:58 PM World Community Grid General prefs: using your defaults 5/11/2012 8:23:58 PM Reading preferences override file 5/11/2012 8:23:58 PM Preferences: 5/11/2012 8:23:58 PM max memory usage when active: 2635.93MB 5/11/2012 8:23:58 PM max memory usage when idle: 7188.90MB 5/11/2012 8:23:58 PM max disk usage: 10.00GB 5/11/2012 8:23:58 PM don't use GPU while active 5/11/2012 8:23:58 PM (to change preferences, visit the web site of an attached project, or select Preferences in the Manager) 5/11/2012 8:23:58 PM Not using a proxy 5/11/2012 8:23:58 PM Suspending computation - initial delay 5/11/2012 8:25:57 PM World Community Grid Restarting task SN2S_q00277_0000005_0436_0 using sn2s version 613 5/11/2012 8:25:57 PM World Community Grid Restarting task cfsw_1224_01224382_0 using cfsw version 605 5/11/2012 8:25:57 PM World Community Grid Restarting task X0930062350039200512221241_0 using hcc1 version 642 5/11/2012 8:25:57 PM World Community Grid Restarting task X0960062350374200601120934_0 using hcc1 version 642 [Edit 1 times, last edit by Former Member at May 13, 2012 4:32:59 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi,
(by now Lawrence could be off to slumberland, but think he's on west coast, so he could still be up painign his brain what to make of this). boincmgr.exe being run away, when in fact it's with boinc.exe the 2 that use the least of BOINC time does happen. For 6.10.58 it's very rare. My thinking is that your Firewall/AV software is having an issue with these and interferes. boincmgr.exe needs free pass over localhost IP 127.0.0.1 on port 31416 to send / receive the so-called RPC signals. If that blocks in your firewall, boincmgr could start sucking up more CPU time and since it runs at normal priority, that you'll notice. This in turn could cause the AV go crazy, but that would show in some pop up or event log message. If the security software is the cause, the remedy is to set exceptions in your AV/firewall not scan these and let them communicate freely. BOINC usually can run full power, and if this is fixed, suggest you try running at 100%. The throttle is only designed for the laptops that often get hot with their poor internal mechanical cooling. Not critical at all: See with the real account name in the message log (you'd like to edit that out of your last post), that BOINC is installed at user level. My personal recommendation is, if it is your right to install on the work machine, to install BOINC as a service (protected application execution). Then BOINC will continue even when no one is logged in to the machine (if it is shared). Could be that your IT does not allow you to do that though. Check them out if they're OK. BTW, you don't need BOINCcmgr. You can exit it and leave the core client boinc.exe running. Let us know --//-- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello sevenveils,
Thanks for posting so quickly. The first thing I notice is no GPU requiring a local preferences override file for GPU control, so 5/11/2012 8:23:58 PM Reading preferences override file means that you are not using the website preferences for this machine, an unnecessary and possibly confusing step unless you have more than 1 computer. You can get rid of this by bringing up BOINC Manager and selecting Advanced - Preferences and then clicking on Clear in the upper right of the popup.I like your powerful Intel i5 cpu with lots of memory. So right at the start, I am thinking that you must have a complicated software environment to bring it to its knees. Brought from work? Does that mean that it is a laptop? If it is a desktop, it should have adequate cooling, so you could increase the cpu usage to 100%. This would let you use Task Manager to see if the WCG projects are running at about 25%. This would mean that with nothing else going on, they are fully using each core. My guess is that more than that is running on your computer. If so, start reducing the number of cores that BOINC can run [75%,50% or 25%] until BOINC can fully utilize the cores when you are not doing anything else. Considering your description, which sounds like a lot of software contention, you probably should never use more than 3 cores [75%]. What does Task Manager show running with a lot of cpu usage besides the WCG projects?  Hint, if WCG projects show less than 25% but nothing else shows, there is probably something running under the Administrator account.If nothing that I have said seems to fit, still cut the number of cores down to 3. You need to make your computer more responsive. BOINC is only a secondary priority. Lawrence |
||
|
|
mikey
Veteran Cruncher Joined: May 10, 2009 Post Count: 826 Status: Offline Project Badges: 
|
Hi! I've been using WCG on a few different machines over the years, and all has been well til I got a Win 7 Pro (64 bit) machine from work. WCG has not agreed with this machine at all. Despite setting WCG to use no more than 45% of CPU time (I have dialed this down further, but it seems to make no difference), if WCG is running along with Thunderbird or Chrome with a lot of windows open, or both, my CPU will periodically flatline at 100% and everything will freeze up for long periods of time, with the WCG processes definitely adding up to well over 45%. It seems to coexist particularly poorly with Thunderbird. Thanks muchly! I will make one small comment and then let the experts get back to you with the stuff I don't know: Boinc does not work like you think it does! When you set Boinc to 45% it tells it to use 45 out of 100 clock ticks, that means that for 45 clock ticks your cpu runs at 100% then Boinc rests for 55 clock ticks, giving an AVERAGE of 45%! It does NOT mean run 45% of EACH clock tick which is what it seems you are wanting. So what you are seeing is what it is supposed to be doing, running at 100% for a period of time then backing off and not running for a chunk of time, then repeating endlessly.   |
||
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges: 
|
I agree with Lawrence et. al. If you reduce the cores used for BOINC to 3 rather than trying to reduce the CPU% used, you will probably see much better performance on the other tasks.
----------------------------------------  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
5/11/2012 8:23:57 PM Processor: 4 GenuineIntel Intel(R) Core(TM) i5 CPU M 560 @ 2.67GHz [Family 6 Model 37 Stepping 5] I'm pretty sure it's a laptop (the M in the processor model). Speaking for myself I've found that in a laptop, temperature-wise, it's better to throttle than to limit the number of cores. |
||
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges:
|
Thanks for the information bono vox. I didn't know that. I must admit that I have only owned 1 laptop in my life and that was a 1995 Compaq.
----------------------------------------  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The root problem [it seems to me] in the second post of sevenveils is querying of the sluggishness of the device, seemingly caused by boincmgr.exe obstruction [normal priority process]. sevenveils needs to come back on the address made of that issue, when he's ready. Fix that and the CPU time eating is probably going away. And the OS being the OS, it spreads the load over the cores, whether BOINC crunching is allowed to use all cores - 1 i.e. boincmgr.exe load is likely moved to the least loaded core.
As to the heat issue on laptops, as bono_vox noted, it's pretty pointless to run n -x cores, as the system detects the highest core thread heat and will throttle down to where it's save [or continues to run too hot]. My octo-lap CPU (Q2670QM), just goes down and down till it system considers it save, so using as the mood goes, TThrottle, for automatic fine grained [general] slowing of BOINC, or ThreadMasterGUI, for fixed, fine grained slowing, but at individual science level [as different science cause different heat generation... whatever specialty functions a science needs most of a CPU] As to mikey159b's, "When you set Boinc to 45% it tells it to use 45 out of 100 clock ticks", I'm sure he meant "wallclock" ticks as the BOINC throttle works on *whole* rounded seconds, not a % of CPU clock cycles per second (wish it was as that is the key issue with BOINC.. the up and down cycling). WCG tech is considering to knock the default standard / power saving setting to 50%, so the cycle run will be 1 second on, 1 second off, and hard research is in progress to implement an all platform refined throttle, and trust me, if Berkeley does not adopt, WCG will run it as it's own tool as an add-on similar to the old UD agent. I'm picturing here that upgrading of BOINC is then not needed and the add-on tool will just work as TThrottle/ThreadMasterGUI torque the science-app cycles, albeit on Win/Mac/Lnx... well that would be dreamland if latter 2 would get their add on app to slow BOINC. So far, the side trail. --//-- |
||
|
|
|