I wonder how I can force Boinc to only download 1 WU at a time for backup projects. I am running Milkyway as a main project for my ATI GPU with PrimeGrid as a backup. Milkyway only allows me to keep 12 WUs max in the queue, which is good for 15-25 min of work. So even a brief server outage can drain the queue. Hence I have PrimeGrid as a backup project. However, every time that kicks in, Boinc immediately downloads 22 PrimeGrid WUs, which is good for 20h of work or so. This is far more than I want to spend on PrimeGrid. That should really only be run when there is absolutely nothing else to do. I want Boinc to return to Milkyway as quickly as possible. So what I want is an option to download only 1 WU at a time from the backup project until the Milkyway server is back up. How can I do this? I am using Boinc 6.10.58.

There is a <fetch_minimal_work> option in cc_config.xml, which causes only one task to be fetched per device, but that would apply to all projects, not only backup projects; and it works not before version 6.11.

Can you lower your cache settings?

Gruß,
Gundolf

---

Computer sind nicht alles im Leben. (Kleiner Scherz)

Thanks, but fetching 1 WU per time for all projects would be unacceptable. I would be out of work almost immediately every time there is a network / server outage...

I do not see any option to set the cache size with Primegrid either. They just inherit the global setting for all projects with no possibility to override that.

I am really stomped on this one...

There isn't any solution to this. Even with the backup project set to backup resource share of zero, it will download a full batch according to your connect to and additional days amount and run that almost exclusively.

Since the amount of tasks at Milkyway is only 6 per CPU core/GPU per host per scheduler contact, you may want to play around with the additional days amount.

---

OK, too bad...

I am quite happy with the cache settings I have now, they have saved my bacon on more than one occasion. So I don't want to change those.

These short Milkyway outages happen every 24-48h from my limited experience so far, so babysitting Boinc is not really an option. I have manually aborted jobs up to now, but I cannot keep that up in the long run.

So I guess the only option I have is to detach from Primegrid and take the down time whenever it comes...

Thanks!