Could anyone explain the difference between these two screenshots, please?

BOINC running as a service
BOINC running as an application

As a service, BOINC.EXE was using between 6% and 13% of CPU cycles (fluctuating on each refresh of task manager). BOINC Manager wouldn't open properly - not even the menu is visible, and it was unresponsive. I had to crash-close it.

If I stopped the service, the manager opened normally, started the cc, and gave all CPU to the project applications, exactly as it should.

This behaviour was fully reproducible - daemon wasted cycles, client-manager worked as normal. Normally, there's no problem - the service daemon uses next-to-no cycles, and the (dual core) apps run at 50% each. This is out-of-envelope behaviour for the host, today only: but like I said, reproducible.

Test environment is awkward - it's a remote server which I have authority to manage using Remote Desktop. It's only a small test and training network, which means I'll be able to reboot and test again overnight - but not in the middle of the working day.

Also, it's a Windows 2003 Small Business Server, which means it absolutely has to run as a Domain Controller - so no upgrade to BOINC v6.2.x. BOINC also has to run as a service, because 99.9% of the time nobody is logged into the console. You'll see I'm running v5.10.13 (my favourite), but I don't remember there being many relevant changes between that and v5.10.45. I'll stick with .13, thank you very much, because the 15 minute connect/reporting interval I've set makes it easier to monitor without the overhead of setting up the whole VPN/RDP management thing.

Why not try 5.10.45 and see if it does the same thing? If it doesn't, there was a fix somewhere between .13 and .45 (it's only 32 iterations, something must've been added, changed, prodded, fixed).

.45 has the option to report immediately (within a minute) if that's something you want it to do...

---

Why not try 5.10.45 and see if it does the same thing? If it doesn't, there was a fix somewhere between .13 and .45 (it's only 32 iterations, something must've been added, changed, prodded, fixed).

.45 has the option to report immediately (within a minute) if that's something you want it to do...

I may try that after working hours this evening. But it's going back to v5.10.13 afterwards. (Did you hear the one about the validate errors at SETI if you use RRI? That's why it's set at a nice safe 15 minutes) (or 14:24, to be pedantic).

It's definitely something about the state the machine's in today, so I'll definitely be rebooting it remotely when I've finished - got some security patches to install while I'm there, I see. So any further tests I should run need to be notified in the next 6 hours or so.

OK, I left it running all day. This is the log:

StartServiceCtrlDispatcher being called.
This may take several seconds. Please wait.
2008-08-07 11:59:16 [---] Starting BOINC client version 5.10.13 for windows_intelx86
2008-08-07 11:59:16 [---] log flags: task, file_xfer, sched_ops
2008-08-07 11:59:16 [---] Libraries: libcurl/7.16.1 OpenSSL/0.9.8e zlib/1.2.3
2008-08-07 11:59:16 [---] Executing as a daemon
2008-08-07 11:59:16 [---] Data directory: C:\Program Files\BOINC
2008-08-07 11:59:16 [---] BOINC is running as a service and as a non-system user.
2008-08-07 11:59:16 [---] No application graphics will be available.
2008-08-07 11:59:16 [Einstein@Home] Found app_info.xml; using anonymous platform
2008-08-07 11:59:16 [SETI@home] Found app_info.xml; using anonymous platform
2008-08-07 11:59:16 [---] Processor: 2 GenuineIntel Intel(R) Xeon(R) CPU 3040 @ 1.86GHz [x86 Family 6 Model 15 Stepping 2]
2008-08-07 11:59:16 [---] Processor features: fpu tsc pae nx sse sse2 mmx
2008-08-07 11:59:16 [---] Memory: 2.00 GB physical, 3.85 GB virtual
2008-08-07 11:59:16 [---] Disk: 25.00 GB total, 7.83 GB free
2008-08-07 11:59:16 [Einstein@Home] URL: http://einstein.phys.uwm.edu/; Computer ID: 889490; location: school; project prefs: school
2008-08-07 11:59:16 [lhcathome] URL: http://lhcathome.cern.ch/lhcathome/; Computer ID: 9654285; location: home; project prefs: default
2008-08-07 11:59:16 [SETI@home] URL: http://setiathome.berkeley.edu/; Computer ID: 3150564; location: school; project prefs: school
2008-08-07 11:59:16 [---] General prefs: from SETI@home (last modified 2008-04-07 16:56:21)
2008-08-07 11:59:16 [---] Host location: school
2008-08-07 11:59:16 [---] General prefs: using separate prefs for school
2008-08-07 11:59:16 [---] Preferences limit memory usage when active to 2047.27MB
2008-08-07 11:59:16 [---] Preferences limit memory usage when idle to 2047.27MB
2008-08-07 11:59:16 [---] Preferences limit disk usage to 7.83GB
2008-08-07 11:59:16 [SETI@home] Restarting task 03jl08ab.31147.72.15.8.48_1 using setiathome_enhanced version 528
2008-08-07 11:59:16 [SETI@home] Starting 17jn08ac.25817.100967.7.8.137_0
2008-08-07 11:59:16 [SETI@home] Starting task 17jn08ac.25817.100967.7.8.137_0 using setiathome_enhanced version 528
2008-08-07 12:09:22 [SETI@home] Sending scheduler request: To report completed tasks
2008-08-07 12:09:22 [SETI@home] Reporting 1 tasks
2008-08-07 12:26:49 [SETI@home] Computation for task 03jl08ab.31147.72.15.8.48_1 finished
2008-08-07 12:26:50 [Einstein@Home] Restarting task h1_0293.40_S5R4__78_S5R4a_0 using einstein_S5R4 version 604
2008-08-07 12:26:52 [SETI@home] [file_xfer] Started upload of file 03jl08ab.31147.72.15.8.48_1_0
2008-08-07 12:27:42 [SETI@home] Computation for task 17jn08ac.25817.100967.7.8.137_0 finished
2008-08-07 12:27:42 [SETI@home] Starting 17jn08ac.25817.100967.7.8.68_1
2008-08-07 12:27:42 [SETI@home] Starting task 17jn08ac.25817.100967.7.8.68_1 using setiathome_enhanced version 528
2008-08-07 12:27:44 [SETI@home] [file_xfer] Started upload of file 17jn08ac.25817.100967.7.8.137_0_0
2008-08-07 12:44:59 [Einstein@Home] Task h1_0293.40_S5R4__78_S5R4a_0 exited with zero status but no 'finished' file
2008-08-07 12:44:59 [Einstein@Home] If this happens repeatedly you may need to reset the project.
2008-08-07 12:45:43 [Einstein@Home] Restarting task h1_0293.40_S5R4__78_S5R4a_0 using einstein_S5R4 version 604
2008-08-07 12:52:27 [SETI@home] Computation for task 17jn08ac.25817.100967.7.8.68_1 finished
2008-08-07 12:52:27 [SETI@home] Starting 04jl08ab.30358.2935.10.8.110_1
2008-08-07 12:52:27 [SETI@home] Starting task 04jl08ab.30358.2935.10.8.110_1 using setiathome_enhanced version 528
2008-08-07 13:43:56 [SETI@home] Task 04jl08ab.30358.2935.10.8.110_1 exited with zero status but no 'finished' file
2008-08-07 13:43:56 [SETI@home] If this happens repeatedly you may need to reset the project.
2008-08-07 13:43:56 [SETI@home] Restarting task 04jl08ab.30358.2935.10.8.110_1 using setiathome_enhanced version 528
2008-08-07 13:43:57 [Einstein@Home] Task h1_0293.40_S5R4__78_S5R4a_0 exited with zero status but no 'finished' file
2008-08-07 13:43:57 [Einstein@Home] If this happens repeatedly you may need to reset the project.
2008-08-07 13:43:57 [Einstein@Home] Restarting task h1_0293.40_S5R4__78_S5R4a_0 using einstein_S5R4 version 604
2008-08-07 14:18:55 [SETI@home] Computation for task 04jl08ab.30358.2935.10.8.110_1 finished
2008-08-07 14:18:55 [SETI@home] Starting 04jl08ab.30358.4980.10.8.66_0
2008-08-07 14:18:55 [SETI@home] Starting task 04jl08ab.30358.4980.10.8.66_0 using setiathome_enhanced version 528
2008-08-07 14:44:03 [SETI@home] Task 04jl08ab.30358.4980.10.8.66_0 exited with zero status but no 'finished' file
2008-08-07 14:44:03 [SETI@home] If this happens repeatedly you may need to reset the project.
2008-08-07 14:44:56 [SETI@home] Restarting task 04jl08ab.30358.4980.10.8.66_0 using setiathome_enhanced version 528
2008-08-07 14:49:18 [SETI@home] Starting 18jn08aa.27205.11115.7.8.168_0
2008-08-07 14:49:18 [SETI@home] Starting task 18jn08aa.27205.11115.7.8.168_0 using setiathome_enhanced version 528
2008-08-07 15:21:29 [SETI@home] Computation for task 18jn08aa.27205.11115.7.8.168_0 finished
2008-08-07 15:21:29 [SETI@home] Starting 18jn08aa.27205.11115.7.8.207_1
2008-08-07 15:21:29 [SETI@home] Starting task 18jn08aa.27205.11115.7.8.207_1 using setiathome_enhanced version 528
2008-08-07 15:44:42 [SETI@home] Task 18jn08aa.27205.11115.7.8.207_1 exited with zero status but no 'finished' file
2008-08-07 15:44:42 [SETI@home] If this happens repeatedly you may need to reset the project.
2008-08-07 15:44:43 [SETI@home] Restarting task 18jn08aa.27205.11115.7.8.207_1 using setiathome_enhanced version 528
2008-08-07 15:45:33 [SETI@home] Task 18jn08aa.27205.11115.7.8.207_1 exited with zero status but no 'finished' file
2008-08-07 15:45:33 [SETI@home] If this happens repeatedly you may need to reset the project.
2008-08-07 15:45:33 [SETI@home] Restarting task 18jn08aa.27205.11115.7.8.207_1 using setiathome_enhanced version 528
2008-08-07 15:50:35 [SETI@home] Computation for task 18jn08aa.27205.11115.7.8.207_1 finished
2008-08-07 15:50:35 [SETI@home] Starting 04jl08ab.2409.13160.15.8.168_0
2008-08-07 15:50:35 [SETI@home] Starting task 04jl08ab.2409.13160.15.8.168_0 using setiathome_enhanced version 528
2008-08-07 15:57:01 [SETI@home] Computation for task 04jl08ab.30358.4980.10.8.66_0 finished
2008-08-07 15:57:01 [SETI@home] Starting 17jn08ad.32708.13571.10.8.246_1
2008-08-07 15:57:01 [SETI@home] Starting task 17jn08ad.32708.13571.10.8.246_1 using setiathome_enhanced version 528
2008-08-07 16:57:09 [Einstein@Home] Resuming task h1_0293.40_S5R4__78_S5R4a_0 using einstein_S5R4 version 604
2008-08-07 17:01:30 [SETI@home] Computation for task 04jl08ab.2409.13160.15.8.168_0 finished
2008-08-07 17:01:30 [SETI@home] Resuming task 17jn08ad.32708.13571.10.8.246_1 using setiathome_enhanced version 528
2008-08-07 17:58:06 [SETI@home] Starting 17jn08ad.32708.21751.10.8.235_0
2008-08-07 17:58:06 [SETI@home] Starting task 17jn08ad.32708.21751.10.8.235_0 using setiathome_enhanced version 528
2008-08-07 18:17:37 [SETI@home] Computation for task 17jn08ad.32708.13571.10.8.246_1 finished
2008-08-07 18:17:37 [SETI@home] Starting 17jn08ae.32037.1708.10.8.7_0
2008-08-07 18:17:37 [SETI@home] Starting task 17jn08ae.32037.1708.10.8.7_0 using setiathome_enhanced version 528
2008-08-07 18:45:33 [Einstein@Home] Task h1_0293.40_S5R4__78_S5R4a_0 exited with zero status but no 'finished' file
2008-08-07 18:45:33 [Einstein@Home] If this happens repeatedly you may need to reset the project.
2008-08-07 19:18:02 [Einstein@Home] Restarting task h1_0293.40_S5R4__78_S5R4a_0 using einstein_S5R4 version 604
2008-08-07 20:06:01 [SETI@home] Computation for task 17jn08ad.32708.21751.10.8.235_0 finished
2008-08-07 20:06:01 [SETI@home] Resuming task 17jn08ae.32037.1708.10.8.7_0 using setiathome_enhanced version 528
2008-08-07 20:18:02 [SETI@home] Starting 17jn08ad.8158.12753.11.8.219_1
2008-08-07 20:18:02 [SETI@home] Starting task 17jn08ad.8158.12753.11.8.219_1 using setiathome_enhanced version 528
2008-08-07 20:45:13 [SETI@home] Task 17jn08ae.32037.1708.10.8.7_0 exited with zero status but no 'finished' file
2008-08-07 20:45:13 [SETI@home] If this happens repeatedly you may need to reset the project.
2008-08-07 20:45:13 [SETI@home] Task 17jn08ad.8158.12753.11.8.219_1 exited with zero status but no 'finished' file
2008-08-07 20:45:13 [SETI@home] If this happens repeatedly you may need to reset the project.
2008-08-07 20:45:13 [SETI@home] Restarting task 17jn08ae.32037.1708.10.8.7_0 using setiathome_enhanced version 528
2008-08-07 20:45:13 [SETI@home] Restarting task 17jn08ad.8158.12753.11.8.219_1 using setiathome_enhanced version 528
2008-08-07 21:09:11 [SETI@home] Computation for task 17jn08ae.32037.1708.10.8.7_0 finished
2008-08-07 21:09:11 [Einstein@Home] Resuming task h1_0293.40_S5R4__78_S5R4a_0 using einstein_S5R4 version 604
2008-08-07 21:45:13 [SETI@home] Task 17jn08ad.8158.12753.11.8.219_1 exited with zero status but no 'finished' file
2008-08-07 21:45:13 [SETI@home] If this happens repeatedly you may need to reset the project.
2008-08-07 21:45:13 [SETI@home] Restarting task 17jn08ad.8158.12753.11.8.219_1 using setiathome_enhanced version 528
2008-08-07 22:06:39 [---] Exit requested by user

Notice: lots of tasks exiting with zero status but no 'finish' file. Lots of tasks finished too, but none of them uploaded.

Then I upgraded to BOINC v5.10.45, to keep Jord happy. All the files uploaded while BOINC was running benchmarks (version change), but as soon as computation started, we were back to app 50%, app 35%, BOINC 15% and the manager froze.

Stopped the service with services.msc, started the manager as an application: 50%, 50%, 0% and the manager is responsive.

Did the security patches and the reboot. Service now running 50, 50, 0 and manager is responsive.

Re-installed v5.10.13. Service runs properly, manager fails to load: BOINC.DLL is missing. Use /a switch to extract, move to folder, loads OK - all seems good. Has already reported tasks uploaded by v5.10.45.

Decide not to meddle any further.
--------------------------------
It <feels> as if something got strangled in the comms layer of the machine - which seems odd, since I had no problem getting into it with remote desktop, over the same network interface.

But the question which is intriguing me is: why did BOINC --daemon behave so differently from BOINCmgr|BOINC? Both v5.10.13 and v5.10.45 behaved the same (not at all to my surprise), so it's a current problem - as I said, it's a domain controller, so we're not going into the v6.2.x sequence.

Ideas?

No ideas? Here it is again:



Again on a server, this time Windows 2000 and not a domain controller.

There has been some discussion at SETI recently about BOINC showing excessive usage, but that has been identified as a side-effect of running with cached task lists running into the hundreds or even thousands. This is my home server, with just a humble P4 Northwood (single core) and usually no more than 5 tasks on board.

I noticed it because BOINC was too busy doing something to respond to my BoincView monitor. BV now reports it running Einstein at 99.32% efficiency, so problem solved.

I think both these cases may relate to the length of time that these small servers may be left running unattended without rebooting. This morning, I finally got round to installing a set of Windows Updates that had been hanging around for, probably, months: it was the reboot which did the trick.

BOINC as a Windoze Update laziness monitor? That's another selling point, alongside recovering stolen laptops.

There has been some discussion at SETI recently about BOINC showing excessive usage, but that has been identified as a side-effect of running with cached task lists running into the hundreds or even thousands. This is my home server, with just a humble P4 Northwood (single core) and usually no more than 5 tasks on board.

No more than 5? Your first screenshots showed quite more than that :D

---

No more than 5? Your first screenshots showed quite more than that :D

Different machine - first was a dual core SBS 2003, second a single core Standard Server 2000.

BOINC (installed on a Windows Server, running as a daemon) is reminding me that it's time to install security updates - again.

This time, I had a copy of Process Explorer handy:



Almost 50% CPU usage in ADVAPI32.DLL ??? What's that about?

This is my home server, Windows 2000, single P4: one Einstein task running, one SETI task waiting to run (in memory), another Einstein ready to run. That's all.

There are definitely many, reliable reports that BOINC Manager gets overwhelmed when the task list approaches 1,000 or 1,500 tasks. That would seem to be inevitable with a BM refresh rate - requesting RPC, parsing, sorting, displaying - of 1 second.

There are also reports - perhaps fewer and less certain - of the core client itself clogging up under even heavier load - 5,000 or 6,000 tasks. Some of that overload has been attributed to the work involved in preparing, formatting and communicating the responses to RPC requests. That work will continue, whichever manager is used - though if BoincTasks can be configured to refresh RPC data less frequently, the demand on the core client will be correspondingly lower.

But in the very rare and specific case I've been documenting in this thread, none of the above applies. 3 tasks, BOINC Manager is never opened (except for testing), and the only RPC requests come from BoincView once every 30 seconds.

It's very improbable, but could an AV security software of some kind be impairing inter process comms in this case?

Highly unlikely. This is a home file/print server only, not used for any interactive tasks (I only access it by RDP): it has a very light-touch AV installed.

As noted lower down the thread, this is a problem I have only seen with service installations: it's a BOINC v5, so I can run the exact same files in Manager/Client mode, and BOINC doesn't waste cycles then. The problem only occurs when there are security updates downloaded and waiting to be installed. Rebooting alone is not sufficient to clear it: installing updates and rebooting invariably does clear it.

Okay, I've had BoincTasks up for 4 hours since upgrading to 0.32 with a refresh to monitor local and remote at 10 seconds intervals. It's used 13 seconds in that 4 hours, running on a quad.

Did you check to see how much BOINC CC time was used servicing those BoincTasks data requests?

As the tasks are read about every second this will stress out the Client quite a bit.
Another problem with the BOINC manager is that it is not multi threaded, probably because the code should be the same over various OS.
This means the code is a lot less efficient on Windows. Where everything relies on messages send from and to different threads, so effectively using more than one core.

http://en.wikipedia.org/wiki/Windows_API

Happened again at the beginning of this month. Since the last show, I've upgraded the BOINC version (to v6.10.34), which means I can't (so easily) re-run the "runs in user mode / gets stuck in service mode" test I did with BOINC v5.

Still gets stuck in advapi32.dll:

Still gets stuck in advapi32.dll

Check the stack on TCP/IP->your computer at port 31416. (Not localhost!)

---

ADVAPI32.DLL contains, among others, registry reasd/write functions. I've had a cpu-usage problem with boinc manager a while ago, where it would re-read all its registry settings every few seconds (see http://boinc.berkeley.edu/dev/forum_thread.php?id=4491#28701). Turned out to be caused bij kaspersky internet security 2010. I fixed that by a uninstall/re-install of Kaspersky.

You might want to check WHAT boinc is doing during high cpu usage by using procmon, which will show things like file i/o, registry operations etc.