The scheduling seems to be querky. If one work unit is running in high priority, the scheduler seems to go throught restarting different task as can be seen by the messages below: Notice that there are just a few seconds between restarts when it should be 60 minutes. This was happening in BY 94 and 95 also. It only happens on machines that have priority running wus.
Hope this helps you find the problem. If I can be of any other assistance please let me know.
office
1881 QMC@HOME 27-12-2010 05:28 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1882 Collatz Conjecture 27-12-2010 05:35 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1883 QMC@HOME 27-12-2010 05:35 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1884 Collatz Conjecture 27-12-2010 05:39 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1885 QMC@HOME 27-12-2010 05:42 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1886 Pirates@Home 27-12-2010 05:45 PM Sending scheduler request: Requested by project.
1887 Pirates@Home 27-12-2010 05:45 PM Not reporting or requesting tasks
1888 Pirates@Home 27-12-2010 05:45 PM Scheduler request completed
1889 Collatz Conjecture 27-12-2010 05:47 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1890 QMC@HOME 27-12-2010 05:47 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1891 Collatz Conjecture 27-12-2010 05:51 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1892 QMC@HOME 27-12-2010 05:54 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1893 Quake-Catcher Network 27-12-2010 05:55 PM Sending scheduler request: To send trickle-up message.
1894 Quake-Catcher Network 27-12-2010 05:55 PM Not reporting or requesting tasks
1895 Quake-Catcher Network 27-12-2010 05:55 PM Scheduler request completed
1896 Collatz Conjecture 27-12-2010 05:59 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1897 QMC@HOME 27-12-2010 05:59 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1898 Collatz Conjecture 27-12-2010 06:04 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1899 World Community Grid 27-12-2010 06:06 PM Restarting task c4cw_target03_002605369_1 using c4cw version 613
1900 Collatz Conjecture 27-12-2010 06:14 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1901 QMC@HOME 27-12-2010 06:14 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1902 Collatz Conjecture 27-12-2010 06:17 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1903 QMC@HOME 27-12-2010 06:19 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1904 Collatz Conjecture 27-12-2010 06:21 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1905 QMC@HOME 27-12-2010 06:22 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1906 World Community Grid 27-12-2010 06:23 PM Restarting task c4cw_target03_002605369_1 using c4cw version 613
1907 QMC@HOME 27-12-2010 06:23 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1908 Quake-Catcher Network 27-12-2010 06:25 PM Sending scheduler request: To send trickle-up message.
1909 Quake-Catcher Network 27-12-2010 06:25 PM Not reporting or requesting tasks
1910 Quake-Catcher Network 27-12-2010 06:25 PM Scheduler request completed
1911 Collatz Conjecture 27-12-2010 06:26 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1912 QMC@HOME 27-12-2010 06:28 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
1913 Collatz Conjecture 27-12-2010 06:38 PM Restarting task collatz_1292225166_420580_1 using collatz version 200
1914 QMC@HOME 27-12-2010 06:38 PM Restarting task qasino_b3lypqz-13_s22test.8549_0 using qasinoAlpha version 501
Quote from: idahofisherman on December 28, 2010, 01:48:08 AM
The scheduling seems to be querky. If one work unit is running in high priority, the scheduler seems to go throught restarting different task as can be seen by the messages below: Notice that there are just a few seconds between restarts when it should be 60 minutes. This was happening in BY 94 and 95 also. It only happens on machines that have priority running wus.
This has nothing to do with BT. It's an issue with some of the BOINC versions. I'd suggest updating to BOINC 6.12.6 at the moment.
Quote from: idahofisherman on December 28, 2010, 01:48:08 AM
The scheduling seems to be querky.
idahofisherman,
Quirky is an understatement--but it's also by design. For information on how task scheduling actually works, I suggest you read this BOINC FAQ article (http://boincfaq.mundayweb.com/index.php?language=1&view=168&sessionID=04e0c395fad815844d992cc2a1c354d7) and the BOINC WIKI detailed description (http://boinc.berkeley.edu/trac/wiki/ClientSched). You'll see there's
a lot more going on than just changing every 60 minutes. And not all projects' types of work units play well together under BOINC's Scheduling Policy. But if you still think something is broken, the BOINC message board (http://boinc.berkeley.edu/dev/forum_forum.php?id=10) is the appropriate place to post your concerns ;)
Best wishes for the new year.
Quote from: idahofisherman on December 28, 2010, 01:48:08 AMNotice that there are just a few seconds between restarts when it should be 60 minutes.
Notice he's saying restarts are every few seconds when it should be 60 minutes. This is a bug that was in MANY BOINC versions and has finally been squashed in the last few betas (keep fingers crossed) :)
From the "Re: BT 0.73" thread:
Quote from: Pepo on September 14, 2010, 01:31:07 PM
When I have to do and the sluggish windows and cursor are hindering me, I'm setting "√ Snooze Gpu". Today, already for the second time, I'm noticing that already hours after snoozing my GPU (when the tasks should have been resumed after the usual one hour), the BOINC symbol on BT's tray icon is still red-blinking. BOINC Manager's tray icon context menu's "Snooze GPU" entry is not checked anymore, a GPU task already runs (also highlighted as such in BT), but BT's tray icon menu still and still says "√ Snooze Gpu"...
I've noticed the same now. Two differences: the red pause circle in the tray icon did not flash, but still kept solid red
and the BT's tray menu did correctly display the GPU snooze time (if it was snoozed) or no snooze time (while running GPU).
Just restarting BT repaired this discrepancy to the normal state.
Quote from: Pepo on January 07, 2011, 04:35:57 PM
Just restarting BT repaired this discrepancy to the normal state.
But after next hour of GPU snooze times out, it happens again.
Quote from: Pepo on January 07, 2011, 05:03:50 PM
Quote from: Pepo on January 07, 2011, 04:35:57 PM
Just restarting BT repaired this discrepancy to the normal state.
But after next hour of GPU snooze times out, it happens again.
On the bug list
Quote from: Pepo on January 07, 2011, 05:03:50 PM
But after next hour of GPU snooze times out, it happens again.
I can't find anything wrong.
Try this again after rebooting the entire computer.
If this still happens and the red dot is stuck, try snoozing and immediately remove the snooze again.
Quote from: fred on January 11, 2011, 07:58:21 PM
Quote from: Pepo on January 07, 2011, 05:03:50 PM
But after next hour of GPU snooze times out, it happens again.
I can't find anything wrong.
Try this again after rebooting the entire computer.
I will afterwards. It will maybe take a couple more days until it happens.
Quote
If this still happens and the red dot is stuck, try snoozing and immediately remove the snooze again.
Does not mater. Until the first complete snooze interval runs over, the red dot behaves correctly. After the interval passes over, the dot remains stuck solid. Just BT's restart helps. (Let's see then after a reboot.)
One more thing
(http://www.efmer.eu/forum_tt/Themes/default/images/post/lamp.gif) maybe: I'll try to find out (maybe with some periodic alarm reminder clock
:D), whether the solid red dot appears exactly at the end of snooze interval, or sooner, or later. Before I've just
noticed (because the screen rendering got again sluggish after GPU crunching got enabled), that "the snooze is apparently over and the dot is again solid red", but do not know the exact moment.
Quote from: Pepo on January 11, 2011, 09:12:15 PM
Quote from: fred on January 11, 2011, 07:58:21 PM
I can't find anything wrong.
Try this again after rebooting the entire computer.
One more thing (http://www.efmer.eu/forum_tt/Themes/default/images/post/lamp.gif) maybe: I'll try to find out (maybe with some periodic alarm reminder clock :D), whether the solid red dot appears exactly at the end of snooze interval, or sooner, or later. Before I've just noticed (because the screen rendering got again sluggish after GPU crunching got enabled), that "the snooze is apparently over and the dot is again solid red", but do not know the exact moment.
Does not happen anymore. I've tried two snooze cycles, but the red dot disappeared.
No reboot yet.
I've interestingly noticed (with B 6.12.10) that after snoozing GPU computation with BOINC Manager, the Manager's
Activity menu goes from my preferences "CPU-prefs + GPU-Always + Netw-prefs" into some weird intermediate state "CPU-prefs + GPU-Always + Netw-suspended" and after a long timeout (few tens of seconds) finally into expected state "CPU-prefs + GPU-suspended + Netw-prefs". Does anybody else see this?
Finally I've brought BT into a state, where GPU is snoozed (for 57:31 now) and the icon is solid white :P For a long time and number of attempts, BT's tray icon copied the state correctly. Now it stays solid white.
New messages prompt won't go away. If I visit the messages page, all of the messages should be marked as "not new". BOINC Tasks keeps flipping to the foreground (yuck, please do NOT do this as it takes the keyboard away from whatever I am doing) - this may be related to the new messages problem.
NOTE that at the moment, I cannot use 0.97 as I get a fatal error with the language file.
Quote from: John McLeod VII on January 17, 2011, 05:27:51 PM
New messages prompt won't go away. If I visit the messages page, all of the messages should be marked as "not new". BOINC Tasks keeps flipping to the foreground (yuck, please do NOT do this as it takes the keyboard away from whatever I am doing) - this may be related to the new messages problem.
I store the last received message time. When a message with a later time stamp appears the icon message should popup.
But sometimes the client keeps refreshing the received time.
But when does BT comes to the foreground.
1) When the Window is open but not on the foreground.
2) Minimized.
3) Hidden in the taskbar.
Quote from: fred on January 17, 2011, 06:11:26 PM
Quote from: John McLeod VII on January 17, 2011, 05:27:51 PM
New messages prompt won't go away. If I visit the messages page, all of the messages should be marked as "not new". BOINC Tasks keeps flipping to the foreground (yuck, please do NOT do this as it takes the keyboard away from whatever I am doing) - this may be related to the new messages problem.
I store the last received message time. When a message with a later time stamp appears the icon message should popup.
But sometimes the client keeps refreshing the received time.
But when does BT comes to the foreground.
1) When the Window is open but not on the foreground.
2) Minimized.
3) Hidden in the taskbar.
1) When the Window is open but not in the foreground.
Not certain about #2 or #3.
Quote from: John McLeod VII on January 17, 2011, 08:21:46 PM
1) When the Window is open but not in the foreground.
It is best to close the window when not in use, less resources are used that way.
For the taskbar popup the main window needs to be selected, and when BT does this and the window is still somewhere open it may be brought to the foreground.