BOINC Manager connection and work issues

Message boards : BOINC Manager : BOINC Manager connection and work issues
Message board moderation

To post messages, you must log in.

AuthorMessage
Aegis Maelstrom

Send message
Joined: 9 Jan 16
Posts: 2
Poland
Message 66726 - Posted: 9 Jan 2016, 10:52:24 UTC
Last modified: 9 Jan 2016, 10:52:53 UTC

Hi there,

I have been crunching BOINC for years and I have installed it on a new, Win 10 64bit system - version 7.6.9 with Virtual Machine. Unfortunately, a number of issues occur.

The major one are problems with the connection of BOINC client with projects. Sometimes it takes long and has hiccups. And today after the system wake I noticed that the tasks are not moving forward (the timecount is not growing) despite of their status "running". I checked the physical files in projects' directories - they have not been updated, and there were almost no new entires in the log, with an exception for a new task of WUProp, info about a possible update and a warning:
New system time (1452329131) < old system time (1452415526); clearing timeouts

This was the last message in the log. I haven't been changing the system time, it seems to be O.K. - however in past something was chenging my date, pushing it 1 day forward.

I tried to restart only the client but nothing was changed. After closing both client and the projects and a restart, the BOINC manager could not connect with the projects and it all seemed to be halted. CPU was barely used all the time, despite of settings to use all cores. After a forced closure of the manager, it returned to life but all the tasks firstly got 0% status (but with hours elapsed kept intact) and then they got an error.

What was surprising, it happened only to the CPU Universe@home tasks. Rosetta, WUProp and a GPU task for PrimeGrid kept on going.

On top of that, BOINC on that machine got some other issues as well. E.g. it did not restart work after a temporary work halt in the proferences settings and the whole machine hardfroze (it was working but it was completely unresponsive). I saw it twice.

Additionally, PrimeGrid GFN tasks crash after 1 sec and BOINC does not recognize Intel GPU on the Intel i7.
All of these may be unrelated or show some issues with the PC or the BOINC client.
ID: 66726 · Report as offensive
Profile Ageless
Volunteer moderator
Project administrator
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 12109
Netherlands
Message 66727 - Posted: 9 Jan 2016, 14:44:47 UTC - in response to Message 66726.  

Several things.
Win 10 64bit system - version 7.6.9 with Virtual Machine.

If you do not need VirtualBox, there's no real need to install it.
If you do have a project requiring VirtualBox, best update to 7.6.22 with VBox, as that has the new 5.0.10 version of VBox included that is fully compatible with Windows 10.

The major one are problems with the connection of BOINC client with projects. Sometimes it takes long and has hiccups.

Communication problems are almost always due to interfering firewalls, anti-virus software and other anti-malware products.

With communication problems I do mean HTTP errors in the messages, not that it takes a little longer for BOINC to communicate with the project. That can have a myriad of causes, that without us knowing what kind of internet connection you have -e.g. cable/DSL/LAN/satellite, how much hangs onto your connection, what is all downloading and taking up bandwidth, where you are physically, how many hops it takes to the projects, if there is anything amiss on the way or at the project in question- or what you are trying to do, would be difficult to diagnose.

And today after the system wake I noticed that the tasks are not moving forward (the timecount is not growing) despite of their status "running". I checked the physical files in projects' directories - they have not been updated, and there were almost no new entires in the log, with an exception for a new task of WUProp, info about a possible update and a warning:
New system time (1452329131) < old system time (1452415526); clearing timeouts

System wake from what? From hibernation or sleep? Or from a cold boot? Opening the lid of a laptop/notebook? Waking up from how long ago?

When closing the lid on a laptop/notebook, or putting the system to sleep or hibernate, BOINC stops calculations. If you don't see tasks continuing to run on wake up from sleep/hibernation, wait a couple of minutes until the system is fully restarted and has loaded everything back into memory. Then exit & restart BOINC. Do know that this wait time can happen with very little CPU usage, when checking system resources do check disk and memory activity.

This was the last message in the log. I haven't been changing the system time, it seems to be O.K. - however in past something was chenging my date, pushing it 1 day forward.

It can also indicate that your CMOS battery is faulty/empty and needs to be replaced.

I tried to restart only the client but nothing was changed. After closing both client and the projects and a restart, the BOINC manager could not connect with the projects and it all seemed to be halted.

Now, why do so difficult and exit & restart the client only? Why not use BOINC Manager->File->Exit BOINC, check "Stop running tasks when exiting the BOINC Manager"->OK for that? At least the manager allows the client to shut down gracefully, not forced as you'd do through Windows Task Manager. By gracefully shutting down the client, it can close down running applications and write their progress to disk.

When you forcefully only close boinc.exe, there's no graceful shutdown and what you describe next is very possible to happen. Although that depends on the sturdiness of the science application. How it reacts to sudden death.

On top of that, BOINC on that machine got some other issues as well. E.g. it did not restart work after a temporary work halt in the proferences settings and the whole machine hardfroze (it was working but it was completely unresponsive). I saw it twice.

Check Windows Event Viewer for problems.
Clean out dust bunnies.
Check with third party applications like Core Temp + GPU-Z or SpeedFan how hot the system becomes when running work.

Additionally, PrimeGrid GFN tasks crash after 1 sec

Report that at Primegrid, it's not something we can fix.

and BOINC does not recognize Intel GPU on the Intel i7.

To be able to do so, it needs drivers that do OpenCL, and that is something you have to install from Intel, not let Windows take care of. Windows -through Windows Update- will install bare drivers, without the CUDA (Nvidia) or OpenCL (Nvidia, AMD and Intel) components required for general purpose GPU work.

So first you tell Windows 10 not to search for and install updates for the Intel GPU, then you go to https://downloadcenter.intel.com/ and download the correct drivers and install those. Reboot the system, restart BOINC and if all is well, your GPU is found.

There is a caveat that some Intel drivers do not work well at some projects, but you best ask there about that. There's also the thing that you need to extend the desktop over two monitors, or use a VGA dummy connector on the Intel GPU if this is a secondary GPU. And BOINC needs use of the <use_all_gpus> flag then.
Jord
Please do not private message me for tech support. Use the forums for that. Tech PMs will be ignored.
ID: 66727 · Report as offensive
Crystal Pellet

Send message
Joined: 23 Apr 12
Posts: 34
Netherlands
Message 66793 - Posted: 12 Jan 2016, 16:40:11 UTC - in response to Message 66726.  

New system time (1452329131) < old system time (1452415526); clearing timeouts

When new system time Sat, 09 Jan 2016 08:45:31 GMT (1452329131) is in the past
compared to old system time (1452415526) Sun, 10 Jan 2016 08:45:26 GMT as it's here the case, BOINC will wait until new time is equal to old time.
ID: 66793 · Report as offensive
Jim1348

Send message
Joined: 8 Nov 10
Posts: 100
United States
Message 66794 - Posted: 12 Jan 2016, 16:49:06 UTC - in response to Message 66726.  

If you have an antivirus, be sure to add the BOINC Program and Data folders to the exclusions list. And I know Universe has been having problems with their estimated run times, which might contribute to it. I would just detach from it for a while and come back later.
ID: 66794 · Report as offensive

Message boards : BOINC Manager : BOINC Manager connection and work issues

Copyright © 2017 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.