Here's my problem.
Years ago I set my work buffer for 3 days work. Single processor machines, not a problem, never has been. My 4p and 2p machines, could be an issue when the work servers go down. The 4p (64 cores total) is lucky to cache 6 hours of work, even then it's not enough to keep all cores working. The 2p (12 cores total) is about a day or so before it's out. I'd set it for 7 days, but I don't want 7 days worth of work in case of a power blip or crash. I currently just have the 2 machines crunching, so I guess I could bump the buffer...
Is this a Rosetta thing where there is a total amount of units sent to a particular machine? Or something on my end I'm missing?
May be unrelated. Sometimes completed units don't want to send out, and it has to be done manually. Mainly noticed it on the 2p, where I may find a couple dozen units "ready to report". For that machine it's common. However, sometimes the 4p will do it, and I've caught it where it only has a couple units left to crunch. I've gotten into the habit of checking on those rigs twice a day, though I shouldn't have to.
Both rigs are running Ubuntu 12.04, and the 4p also has TEAR'S optimized BOINC client to keep the cores fed. Both rigs are running stock clock Opterons, on Super-Micro boards.
Years ago I set my work buffer for 3 days work. Single processor machines, not a problem, never has been. My 4p and 2p machines, could be an issue when the work servers go down. The 4p (64 cores total) is lucky to cache 6 hours of work, even then it's not enough to keep all cores working. The 2p (12 cores total) is about a day or so before it's out. I'd set it for 7 days, but I don't want 7 days worth of work in case of a power blip or crash. I currently just have the 2 machines crunching, so I guess I could bump the buffer...
Is this a Rosetta thing where there is a total amount of units sent to a particular machine? Or something on my end I'm missing?
May be unrelated. Sometimes completed units don't want to send out, and it has to be done manually. Mainly noticed it on the 2p, where I may find a couple dozen units "ready to report". For that machine it's common. However, sometimes the 4p will do it, and I've caught it where it only has a couple units left to crunch. I've gotten into the habit of checking on those rigs twice a day, though I shouldn't have to.
Both rigs are running Ubuntu 12.04, and the 4p also has TEAR'S optimized BOINC client to keep the cores fed. Both rigs are running stock clock Opterons, on Super-Micro boards.
