| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Africa Rainfall Project
Thread: ARP is not able to save real worktime if i restart my PC |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 18
|
|
| Author |
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12439 Status: Offline Project Badges: 
|
If you only run half of your threads with ARP and the rest with MCM/OPN, as is recommended, then suspending shouldn't be too onerous. Generally, the more threads you have, the faster the machine and the shorter time between checkpoints.
When you know that you need to reboot, get some extra units of MCM/OPN and then suspend your ARP units as they reach a checkpoint. That way you will keep all threads running until you can reboot. Once you have rebooted, you can resume your ARP units. Mike |
||
|
|
BobbyB
Veteran Cruncher Canada Joined: Apr 25, 2020 Post Count: 609 Status: Offline Project Badges: 
|
MJH333:
----------------------------------------I suspend each ARP task just after a checkpoint (e.g. just after 12.5%, 25% etc). I presume you do this manually. Right?I don't reboot my Linux machines very often, but when I need to (e.g. following a Kernel update) I don't even bother with updates any more. They work, are not broken so why fix them. I told them to not update but both Ubuntu and Windows still do ask. At least Ubuntu does nothing without intervention. Windows just updates and boots no questions asked. These 4 machines just do WCG 24/7Option 1: Depending on the machine (how much work is cached) you can increase the number of days of work to cache. I believe the default is 0.2 days. Do a project update and let it get lots of WUs then set "No new tasks" in Boincmgr. 1a): Now let APR finish up, suspend the whole project (after checking how much work will be lost) and boot to Windows.... or just boot. Yes it's all manual and if you get distracted and forget then you may have a dormant machine until you remember. I got caught. Option 2: Go to your WCG profile for that machine and turn off ARP. Go to 1a above. Why the need to boot to Windows every few days? I have 1 such machine and only boot to Windows to keep it up to date but let the every few days to extend to about 2 weeks. Am looking to split the machine into 2. I use option 2. This way if I forget then I still get work done but no APR. [Edit 3 times, last edit by BobbyB at Aug 26, 2021 3:22:53 PM] |
||
|
|
PMH_UK
Veteran Cruncher UK Joined: Apr 26, 2007 Post Count: 774 Status: Recently Active Project Badges: 
|
BOINCTasks can be told to suspend a task at next checkpoint.
----------------------------------------As long as there are enough other tasks that do checkpoint you can set then return later to do restart. Paul.
Paul.
|
||
|
|
nyanthiss
Cruncher Joined: Nov 23, 2012 Post Count: 15 Status: Offline Project Badges: 
|
Hey everyone,
----------------------------------------Thanks for the all the tips. I was aware of the suspending just after checkpoint, i have been using it when i possible. Thanks for the tip about BOINCTasks, i will take a look at it. I have been running other projects (MCM / OPN) on my other machines for quite some time, and those machines can't really run ARP (because memory or because ARM), so i'll keep running those as they are, and ARP running on my laptop at night. This is not an "i'm leaving this project if this isn't fixed" -issue for me, it's just a (mildly annoying) inconvenience. The laptop is 4C/8T and i'm running 4 ARP tasks, i'll test with 4 more MCMs (have too many OPN already).
Intel Xeon E3-1231 v3
AMD A10 7800 AMD Ryzen 5 3500U AMD Ryzen 1700X AMD Ryzen 5900X 2x RaspberryPi, 1x Odroid |
||
|
|
MJH333
Senior Cruncher England Joined: Apr 3, 2021 Post Count: 268 Status: Offline Project Badges: 
|
Bobby - yes I was suspending manually.
Paul - many thanks for the tip about BoincTasks. I didn't realise you could do this because I had the Settings set so that the Checkpoints did not show. That's a real help. Cheers, Mark |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1679 Status: Offline Project Badges: 
|
On my side, I do use BoincTasks since years and the "suspend at checkpoint" functionality is really convenient in particular for ARP1.
----------------------------------------However BoincTasks shall run in the background since it will suspend the task itself. Initially, I made the failure to set "suspend at checkpoint" and to close BoincTasks; thinking that BoincTasks forward the command to the corresponding boincmanager instance: it is not the case! BoincTasks shall be kept running and it will monitor the running tasks at each couple of seconds and, as soon as the checkpoint is reached, it will suspend the task. For people operating several crunching machines, BoincTasks is the most valuable tool making possible to operate the machines headless since BoincTasks provides a fully functional graphical user interface from any Windows machine (BoincTasks) or Linux machine (BoincTasks JS; otherwise BoincTasks can also run in Wine). Cheers, Yves --- To download BoincTasks: https://efmer.com/boinctasks/ |
||
|
|
PMH_UK
Veteran Cruncher UK Joined: Apr 26, 2007 Post Count: 774 Status: Recently Active Project Badges:
|
BOINCTasks JS does not support suspend at checkpoint currently and has other limitations.
----------------------------------------I tried it having lost my last Windows PC but plan to install WINE & BOINCTasks. Paul.
Paul.
|
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1323 Status: Offline Project Badges: 
|
On my side, I do use BoincTasks since years and the "suspend at checkpoint" functionality is really convenient in particular for ARP1. 'Suspend at checkpoint' is working fine and suspend takes place seconds to minutes after the checkpoint is written. I discovered only one circumstance where it's not working. When you add or remove one or more computers from your list and have tasks marked for suspending, the 'suspend at checkpoint' is gone after the list of computers is reloaded. You have to set the mark(s) again. |
||
|
|
|