Info | Message |
---|---|
1) Message boards : Number crunching : no credit awarded?
Message 68517 Posted 1 Mar 2023 by Brummig |
When you hold the door open for someone, it's only polite for that person to say "thank you". You can't spend it, can't eat it and can't take it with you, but if someone doesn't take the trouble to thank you, then the next time around you'll likely be letting the door slam shut in their face. That's why the door to my computer is currently shut to CPDN, and those paying customers can go elsewhere. And since they are paying, how about them paying for the expensive electricity that powers all those BOINC hosts? |
2) Message boards : Number crunching : no credit awarded?
Message 68248 Posted 10 Feb 2023 by Brummig |
Then I wouldn't have made any progress :) |
3) Message boards : Number crunching : no credit awarded?
Message 68244 Posted 10 Feb 2023 by Brummig |
Jumping back to a checkpoint is usually very obvious (representing many hours of work lost), and I didn't notice that happening without obvious cause (such as a power failure or manual suspension). I wonder if it's something to do with hibernation causing a small jump back every evening, but not one that's big enough to notice on the progress? The longer a task runs, the more that would multiply up. I have just taken a quick scan through the Universe@Home results for the host, and I can't see any sign that the virtual machine started running faster when the last CPDN task finished (unfortunately I can't go back very far, and other projects have an even shorter results record). |
4) Message boards : Number crunching : no credit awarded?
Message 68236 Posted 9 Feb 2023 by Brummig |
It's a good thought, Dave, though they did in fact use about 40% more CPU time (interestingly, the final reported time for these two tasks is not the same as the time up to the last reported trickle). But that doesn't account for the massive increase in overall time taken, so maybe something was just pushed over the edge by these tasks (whatever that edge might be), and caused a significant long-term performance drop? For example, maybe there was an increase in disk swapping that had a big impact on performance over a long time, but which wasn't enough day-to-day for me to notice whilst I was working (I might just grumble about a slow web page load, or a slow application start, and carry on with life as normal because it wasn't sufficiently annoying). |
5) Message boards : Number crunching : no credit awarded?
Message 68231 Posted 9 Feb 2023 by Brummig |
Your guesses are all wrong. The host is a desktop machine used for work, work that is mostly not CPU intensive, as per the original use case intent of BOINC. Like any other desktop machine I do hibernate it at the end of each working day. Hibernation does mean it takes longer for tasks to complete than they would if the machine was on 24/7, but it doesn't explain why these two tasks stand out compared to other tasks of the same type run on the same host. The "Suspended CPDN Monitor - Suspend request from BOINC..." messages are generated when the machine hibernates, and there are so many of them simply because these tasks took such a long time to complete (BTW, it doesn't aid readablity to copy and paste a large number of identical log messages into a forum post). I do throttle BOINC (latterly with TThrottle) to reduce fan noise and processor temperature, which will also slow progress. However, again that doesn't explain why these two tasks stand out compared to other tasks of the same type run on the same host. Since this is machine is used daily for work, I don't tolerate disk thrashing for very long. There were a few issues with disk thrashing right at the start of the run of these two tasks that I had to manage, but Googling revealed there was a problem with WSL2 virtual memory management, and indeed the problem disappeared (presumably with a Windows update). Also I do keep an eye on tasks during work breaks, so I know pretty quickly if things are less obviously going wrong. Use of WSL2 may slow tasks compared to native, but again that doesn't explain why these two tasks stand out compared to other tasks of the same type run on the same host. Also, it would be unwise to just assume WSL2 is hopelessly slow, as Universe@Home tasks run several times faster on WSL2 than they do on the WSL2 host. If these tasks had behaved as other similar HadAM4 tasks run on this host, they would have finished many months ago. |
6) Message boards : Number crunching : no credit awarded?
Message 68226 Posted 8 Feb 2023 by Brummig |
I don't know why these HadAM4 at N216 resolution v8.52 tasks took so long (and I got them on the rebound because the both of them had timed out after a year). Earlier ones crunched by the same host took much less time. |
7) Message boards : Number crunching : no credit awarded?
Message 68223 Posted 8 Feb 2023 by Brummig |
The second of my two massively long-running HadAM 4 tasks finished on Saturday, uploaded, and reported in. Once again the trickle has failed to appear on the task page (https://www.cpdn.org/result.php?resultid=22197506), and no credit has been awarded. I'll be setting No New Tasks on CPDN until I see the credit and trickles, and another project that does do something with the results I return can have the CPU time and donated electricty. |
8) Message boards : Number crunching : no credit awarded?
Message 68167 Posted 1 Feb 2023 by Brummig |
Just to put some meat on the bones, the task I think Dave is referring to that completed on the 27th January is one of mine (22136276) that completed a year ago, not in 2023, so yes, that got fully credited. Since then two immensely long-running HadAM4 tasks have been running on that host, both slowly gaining credit throughout 2022. These started at the same time, but I had to pause one of them (22197506) for a few weeks due to virtual memory issues (a known WSL2 bug, I suspect, as the problem quietly went away). However, it has been slowly aquiriing the same credit as the other, a few weeks behind. They currently both have the exact same credit, even though the front-runner (22197505) completed on the 16th January 2023 and the other (22197506) is still running. Credit didn't go up after completion, and no extra credit appeared on the aggregation sites. I estimate 22197506 will complete within a week of this post, possibly on Friday. |
9) Message boards : Number crunching : no credit awarded?
Message 68150 Posted 31 Jan 2023 by Brummig |
Firstly I'm looking at my results page. Secondly I'm looking at the aggregation sites (and yes, I have "Do you consent to exporting your data to BOINC statistics aggregation Web sites?" ticked, and have had for a very long time). The last change in credit was at the beginning of December, despite a task completing without error mid January. |
10) Message boards : Number crunching : no credit awarded?
Message 68143 Posted 31 Jan 2023 by Brummig |
Credit still hasn't been granted. |
11) Message boards : Number crunching : no credit awarded?
Message 68030 Posted 25 Jan 2023 by Brummig |
A HadAM4 task one of my hosts was running that was marked as completed on 16 Jan 2023 still has not appeared as a credit on the stats websites. It looks like the final trickle didn't receive any credit: https://www.cpdn.org/result.php?resultid=22197505 |
12) Message boards : Number crunching : How to Prevent OpenIFS Download
Message 68002 Posted 23 Jan 2023 by Brummig |
OK, looks like I've no choice but to opt for No New Tasks (Edit: Or try -1 :)). And yes, it looks like the repo for Ubuntu 20.04 LTS needs updating. |
13) Message boards : Number crunching : How to Prevent OpenIFS Download
Message 67996 Posted 23 Jan 2023 by Brummig |
Thank you for your quick replies. Dave Jackson spotted the problem. At one point, whilst I was using the computer, BOINC quietly downloaded and ran four OpenIFS tasks simultaneously, plus the two hadam4 tasks it was already running, which was hilarious. At the moment the host is still crunching on one hadam4 (along with other, much less memory-intensive non CPDN tasks), so one OpenIFS is way too many. Is it valid to set <max_concurrent> for each of the OpenIFS apps to 0 (I can set it to 1 and try again once the hadam4 has completed)? The official manual doesn't say. |
14) Message boards : Number crunching : How to Prevent OpenIFS Download
Message 67987 Posted 23 Jan 2023 by Brummig |
Is it possible to stop my host from downloading OpenIFS tasks, other than by setting No New Tasks for CPDN? The virtual memory use (disk thrashing) brings my host almost to a standstill, and even if I let it run the trickles don't upload. |
15) Message boards : Number crunching : Why no credit in Statistics?
Message 58709 Posted 6 Sep 2018 by Brummig |
Hooray! This sound came to mind when I read Les' comment: https://www.youtube.com/watch?v=wO9q4H49cGA |
16) Message boards : Number crunching : Lots of pending WUs
Message 58687 Posted 3 Sep 2018 by Brummig |
If the script is running, why have I received no credit for the trickles my host uploaded on the 15th and 28th August? And if the script always starts from the beginning, why hasn't the credit I lost after the crash been restored? This flat-line is shown both on the CPDN website and on BOINCstats. Are you saying the script is running, but not doing anything useful? |
17) Message boards : Number crunching : Zero / Negative Credit
Message 58472 Posted 24 Jul 2018 by Brummig |
Thank you. I had seen that post a while back, but obviously I don't know the details of how CPDN works, what data was lost, temporarily or forever, or why there is no usable backup, and it didn't answer my questions. And there appears to be a strong reluctance to create a single place for people to look to find out what is the current status and if the problem they are experiencing might be known about and being worked on. Hence the explosion of posts asking what happened and reporting problems that may or may not be related to problems raised in other threads. This entire thread itself would not exist if such a status page existed. |
18) Message boards : Number crunching : Zero / Negative Credit
Message 58470 Posted 24 Jul 2018 by Brummig |
Yes, I looked there, ordering both newest first and oldest first. Either way, the third post says nothing about what has happened to credit from trickles. |
19) Message boards : Number crunching : Zero / Negative Credit
Message 58468 Posted 24 Jul 2018 by Brummig |
OK, so far I've found three News and Announcements threads, none of which tells me in the third post down what happened to credit from trickles, no matter which way I order them. What would be really helpful is a thread here https://www.cpdn.org/cpdnboinc/forum_forum.php?id=71 keeping people (who are almost certainly unaware of how CPDN works "under the hood") up to date with what appears to be working and what is broken following the outage. It could include links to discussion threads. As it is, anyone finding they have an issue is faced with trawling through an increasing number of different threads on multiple fora trying to find out if they have a new problem, or if they just need to be patient (and preferably for how long). Eg:
Trickles: Working. Final Upload: Working. ... Credit reporting on CPDN: Broken, with credit from trickles before the outage temporarily lost. Unlikely to be fixed for at least two weeks. See <url>. Credit export: Working, but with bad data (see above). |
20) Message boards : Number crunching : Zero / Negative Credit
Message 58466 Posted 24 Jul 2018 by Brummig |
I've not seen a task complete on zero credit, but I've never had a task complete in under a week. Since completion tended to be a bit of event, I would always look on the website to see what credit I had received (not knowing, until now, that the credit wasn't in fact up-to-date). So what you say would make sense, Les, but only if the credit from all trickles uploaded before the outage was wiped out by whatever was done during outage (the task concerned spanned the outage period). |
©2024 cpdn.org