Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 17
|
![]() |
Author |
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I just found all my WCG projects running in "high priority", because there were at least a hundred too many work units. I keep a default buffer size of 0.1 + 0.5 days, but the scheduler must have gone berserk. So I aborted several days worth, in order to meet the deadline. I still have over three days of work left. This is on a Ryzen 1700 running Ubuntu 18.04.
What happened? I updated from BOINC 7.14.1 to 7.16.3 about a day ago, but that did not change the buffer size. |
||
|
UBT - JohnR
Cruncher Joined: Apr 30, 2006 Post Count: 35 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
7.16.3 Development version
(MAY BE UNSTABLE - USE ONLY FOR TESTING) Ask your question here - https://boinc.berkeley.edu/dev/ |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
7.16.3 Development version (MAY BE UNSTABLE - USE ONLY FOR TESTING) Possibly, but I have upgraded four other Ubuntu machines, and one Windows machine with no problems. Sometimes the WCG Settings get reset randomly I have noticed, and expect that could be it. [Edit 2 times, last edit by Jim1348 at Oct 5, 2019 11:27:15 AM] |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
It is not exactly a bug, but a feature of 7.16.3 relating to how they have changed the BOINC scheduler.
There is a discussion on it at Einstein: https://einsteinathome.org/content/discussion...nown-o2md1#comment-173743 I find that it affects some projects more than others. I don't notice any effect in some cases. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
If you insist on running a TEST client, set your device profile, custom section to limit the number of tasks to assign on a per-science basis.
To revert to the last stable, 7.14.2, go to https://boinc.berkeley.edu and click the Download BOINC button to get it as appropriate for your operating system. |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
You don't get it. It is a permanent feature, not a bug.
----------------------------------------Maybe someone will set a limit though, either in the BOINC client, or else WCG? After enough complaints, I expect someone will do something. [Edit 1 times, last edit by Jim1348 at Oct 8, 2019 10:13:36 AM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
And still it's a TEST client, nothing set in stone yet.
----------------------------------------The scheduler at one time was set to fetch work when a user manually operated the update button... let's not have a connection go to waste even if the buffer is full. If work in automated fashion keeps coming it's likely because the task duration estimation is wrong. Check the DCF which normally should be near 1.000000. Maybe a berserk benchmark is driving that. Anyway, strange how this rampant work fetching is not a topic on the developer's mailing list v.v. 7.16.3. If it's a concern it should be flagged there. [Edit 1 times, last edit by Former Member at Oct 8, 2019 11:28:47 AM] |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
The estimates for the WCG work units look OK to me, but I am still picking up too much work. I had cut it back to six days, but now it is up to ten days, so I have set NNW.
It doesn't seem to be a problem on various other projects I run. Someone needs to go back to the drawing boards somewhere. |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I would add now that I have three Ubuntu 18.04 machines with BOINC 7.16.3 on WCG, all on the same "location". Two of them exhibit the problem, and just keep downloading until they reach the 10 day limit even though I have the buffer set on one to 0.10 + 0.50 days (the default) and the other one to even less, 0.05 + 0.10 days.
----------------------------------------However, the third machine is working properly now, after downloading five days of work (too much). But at least it is not requesting any more. My guess is that something in the PPA version changed, even though they all show 7.16.3. I will uninstall later and try again. Maybe it is fixed. (Or else, maybe WCG is making changes on their end?). Also, there is the possibility of picking up settings from other BOINC projects. They all interact somewhat. [Edit 2 times, last edit by Jim1348 at Oct 8, 2019 6:47:59 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'm running 7.16.3 from the PPA on 5 systems for a week now and ALL are behaving as expected all on Ubuntu 19.04. However, if they start becoming "flaky" I'll turn on sched_op_debug and work_fetch_debug to document.
|
||
|
|
![]() |