I am running on a 9+ year old system, and it is the only one I have available at this time. It has a pretty fast (2.8ghz) processor, but extremely limited memory (1G). I use this machine almost 100% of my waking hours.
The process einstein_s6Bucket_1.01_windows__intel86_SSE2.exe, as well as others (I believe), but only with the Einstein project, refuse to fully unload from memory when computer goes active. It uses (at the moment) 92.94mb of memory. I don't care that It may be swap space, as my resources are so severely limited, it MUST completely unload. And yes, I have configured it to fully unload and repeatedly double checked that it is indeed set that way.
Killing the process results in a "computation error." All time invested seems to be totally lost.
REPEAT: this is only the case, and only an issue that I have with Einstein (SETI & Milky Way don't seem to have this problem).
Please tell me this is a setting issue, AND how to address it. If this can not be addressed, I will have to leave the project, and I really don't want to have to do that.
PS: Sorry if this has been addressed, but I don't have time to repeatedly search and/or read so many pages of posts.
Doh! Just me!
Copyright © 2024 Einstein@Home. All rights reserved.
Processes refuse to unload when computer in use
)
Hi!
If you go to : http://einstein.phys.uwm.edu/prefs.php?subset=global you will find a setting "Leave tasks in memory while suspended?
Suspended tasks will consume swap space if 'yes'" . You will want to set this to "no".
However, this setting is in theory controlling this behavior for ALL of your projects, so this would not explain why your other projects are behaving differently already. It's still a good idea to check thi setting. Note that this setting can be made for different "venues" or classes of hosts. Each host in BOINC is associated to one "venue", see the details for your PC to look up whcih venue it is assigned to.
HBE
RE: Note that this setting
)
In addition, you have the capability to set a preference item for this issue locally, and if you have done so that value takes precedence over anything you have set on the web pages Bikeman pointed you to.
Assuming you are using boincmgr with a user interface similar to that on my copy of 6.12.34, you can check this point by
Boincmgr|Tools|Computing preferences|disk and memory usage|memory usage
and verify that the item "Leave applications in memory when suspended" is NOT checked. Be sure to "cancel" as you leave the preferences dialog, just to be sure you have not newly set an over-riding preference locally.
As previously stated, this does not seem to fit your stated symptoms fully, but it is definitely one way to have a science application fail to release memory.
Thanks folks, I do appreciate
)
Thanks folks, I do appreciate the time and effort you've taken to reply.
Unfortunately, these are both settings that I have found and had already set to achieve stated goals ... other than "venues." Can venues override local settings AND web-based settings?
Does anyone know if this is a known bug? Does the error from killing the process force a restart of the process, wasting all work already performed?
Doh! Just me!
RE: Unfortunately, these
)
The venues matter is not something separate from web-based settings, but rather an aspect of them. Just to make things a little confusing, the same subdivision is called a venue in some references, a location in others, and no name at all in yet others.
Specifically, on the computing preferences or the Einstein@home preferences web page of your account at, for example, Einstein@home, the default/home/school/work distinction is what we mean by venue or location. If you click the "switch view" link near the top of the page, you'll see values for the four cases arranged in easily compared vertical columns. There is a column headed "location" in the "your computers" list for any project, and if you click on "details" for a particular computer, you'll see the current one for that computer listed in a windows scrolling control labelled "location" quite near the bottom.
Which don't matter in your case, as you report that you've actioned a preference in your copy of boincmgr, which takes precedence over anything trickling down from the web.
And just FYI, I have no
)
And just FYI, I have no separate location/venue configs, as the button reads, create, not update, on those pages. Nor, if I recall correctly, is removal from memory an option at that level? Doesn't matter, not the problem.
I have suspended Einstein, pending resolution of this issue.
Doh! Just me!
Well, it looks like admins
)
Well, it looks like admins and/or the community has not been able to help, so I guess I'll have to say goodbye until I get a computer with some resources to burn.
That's a shame too, as this project (at least) seems to be getting positive results. If anyone who can see my email address wants to send me an update if/when this is addressed, that would be fine. Take care!
Doh! Just me!