Hi
I'm setting up a read-only netboot volume, that will let me boot up any machine to a pre-configured system that has BOINC running. I have everything set up the way I want it. But the next step before I turn this volume into a netboot image, is to delete all current in progress, downloading, and completely work units. That way, when a computer boots this system, it will be starting clean. Otherwise every time I boot a computer, its going to spend hours crunching the same old expired work units before it goes and downloads more.

Also, a second question related to running off a read-only volume. To get the most of of the time spent crunching (considering the fact that when you reboot, any partial result will be lost), is there a way I can make BOINC never switch to a new project mid-work-unit? I have 6 projects and I want them to mix up which it does, but once it gets started on one unit, I want it to stick with it until it's done (or many if it's doing multiple at once)

I probably won't know the answer, but what was the question intended in your 1st post?

I think that the answer to your 2nd question, is to set the Switch between interval to longer than the longest wu will take to complete. The default is 60 minutes.
But all of this breaks several reasons behind the structure of BOINC.

You'll get better answers when Europe and American wake up.

The image should NOT have BOINC running. A script that runs on each machine at boot time should start BOINC.

Huh?

Note that I'm not doing this to set up full time machines that I own. I'm doing this so that when I do get an extra computer around, whether it be one I'm selling myself or something I'm repairing for someone else, I can boot this image and run boinc without having to actually install (and then uninstall) anything on the hard drive of the machine.

I was actually able to reset all the projects and quickly quit and it seems to have left the Boinc manager in just the state I want it to be in. I tried setting the time for each project to 999 minutes but it seems to ignore that value and switch after just a few minutes . . . I'm not sure how many exactly, less than an hour.

Starting with BOINC 6.6.10, there is an option named <abort_jobs_on_exit>:

If 1, abort jobs and update projects when client exits. Useful on grids where client gets wiped after each run. New in 6.6.10

More information to cc_config you can find here.

Gruß,
Gundolf

---

Computer sind nicht alles im Leben. (Kleiner Scherz)

Sadly the newest version for OS X is 6.2

Sadly the newest version for OS X is 6.2

Quote from the All versions download page:

Mac OS X
Version 10.3.9+
6.6.17 Development version (standard GUI)
(MAY BE UNSTABLE - USE ONLY FOR TESTING)

Gruß,
Gundolf

---

Computer sind nicht alles im Leben. (Kleiner Scherz)