![]() |
[QUOTE=swl551;332182]Using WorkToDo as a component of the detection should resolve the issue.
Laurv: I'll have a release in a day or two for you to test.[/QUOTE] Thanks. Related to using worktodo, you can use better the result file, which will change more often in size, or use both, as I commented before (I am thinking of the situation when worktodo does not change in size, only its first line changes, from "bla-bla,71,73" to "bla-bla,72,73", but the result file changes in size too, as a new line is added to it), beside of the method with checkpoints which is better for long assignments. edit: and a bit of nitpicking: after a successful manual report, the "next upload" bar should go to zero (it makes no sense to report again after 5 minutes, if I got impatient and clicked the button just 5 minutes before the bar reached the end of the track...) @Xyzzy: how about here, in "software"? yafu has it in factoring project, but misfit is not about factoring, is a software tool to help with mfaktc/o and cudalucas, so either here, or in gpu-whatever (just a suggestion, other idea welcomed too). |
[QUOTE=LaurV;332185]
@Xyzzy: how about here, in "software"? yafu has it in factoring project, but misfit is not about factoring, is a software tool to help with mfaktc/o and cudalucas, so either here, or in gpu-whatever (just a suggestion, other idea welcomed too).[/QUOTE] +1 |
2.4.7 beta 1
Stalled Process detection now considers the LastWriteTime of the following
1. The latest checkpoint file (if any) 2. WorkToDo.txt 3. Results.txt When [B]all three[/B] stall out then the instance is declared stalled. The side effect of this is that [I]work balancing, work adding/draining staged work and results uploading[/I] can [U]temporarily[/U] mask the fact that a process is actually stalled. Therefore it is recommended you stick with the stalled detection defaults of [B]Minutes between checks=5; Number of times files detected stalled=3; [/B] See LaurV's [B]mind blowing[/B] writeup and stunning graphic to understand the motivation for this change. Although you probably didn't have the issue LaurV describes the code really needs to be tested by persons other than Laurv to ensure I did not break any existing functionality (hey that's why we beta test!) Thanks Scott My formum FTP account is not allowing my login, so for now get from skydrive [URL]http://sdrv.ms/169QpcY[/URL] |
2.4.7 beta 2
I removed the logic detecting changes to WorkToDo as part of the stalled detection process. WorkToDo can be changed by many of MISFIT's native functions which could mask an actual stalled process. On the other hand Results.txt is rarely touched by MISFIT and therefore is less likely to prolong the detection of a stalled process.
It is still recommended you stick with the stalled detection defaults of: [B]Minutes between checks=5; Number of times files detected stalled=3;[/B] Get from [URL]http://mersenneforum.org/misfit/[/URL] |
Tested beta1 for a while, used pause/break key to simulate a stall, which was properly detected and properly resumed when I un'stalled it. Let it run for about two hours, no false alarm. Deleted. "Da capo al fine" with beta2. I hope you don't do a beta3 in 35 minutes :razz:
|
[QUOTE=LaurV;332262]Tested beta1 for a while, used pause/break key to simulate a stall, which was properly detected and properly resumed when I un'stalled it. Let it run for about two hours, no false alarm. Deleted. "Da capo al fine" with beta2. I hope you don't do a beta3 in 35 minutes :razz:[/QUOTE]
No, I think we're done on this issue, unless someone reports bad things! |
[QUOTE=LaurV;332262]Tested beta1 for a while, used pause/break key to simulate a stall, which was properly detected and properly resumed when I un'stalled it. Let it run for about two hours, no false alarm. Deleted. "Da capo al fine" with beta2. I hope you don't do a beta3 in 35 minutes :razz:[/QUOTE]
On another note: Has anyone mentioned your avatar photo is disturbing. Something about it makes me uneasy. I try to avoid looking at it... |
[QUOTE=swl551;332269]On another note: Has anyone mentioned your avatar photo is disturbing. Something about it makes me uneasy. I try to avoid looking at it...[/QUOTE]
That's. strange... Really? :loco: |
[QUOTE=kracker;332273]That's. strange... Really?
:loco:[/QUOTE] Well if you find that photo soothing then mabye you are strange :smile: I KNOW I AM NORMAL. :no: |
Great; I bumped my checkpoint frequency back up to 10 minutes. I was also having the false stall warnings when working the short-duration 332M exponents.
|
[QUOTE=swl551;332274]Well if you find that photo soothing then mabye you are strange :smile: I KNOW I AM NORMAL. :no:[/QUOTE]
It's probably because the "S" is reversed. |
| All times are UTC. The time now is 08:19. |
Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.