It used to be that BOINCMGR trying to talk to the core client and hanging because of blockage by security software or the system to be plain busy that it caused to suck so much juice that the science was unable to tell the core client it's alive. It does so every thirty seconds. Don't know the details, but it's been subject of discussion for years at the Berkeley developers. Get rid of the heartbeat stuff and maybe just monitor the PIDs might be a solution. Long as the science knows the seeding PID still exists and the CC knowing the science app PID is still there assume it's still running, but then they could be running till max time exceed and actually be dead, stuck, looping. Really have not been following, but a recurring nag pain it is, after all these years.

Sek,

I'm sure that Kevin and the rest of the team will do an excellent job, as usual. I was concerned about the "default" behaviour being appropriate for users who don't want to twiddle the knobs and to still give the big crunchers the chance to contribute to the maximum. I guess I'm also just a little concerned about the overhead of qualifying a new BOINC level for the project dependency and the delay that that might cause to the science.

I should, perhaps, also comment that none of my systems showed any adverse behaviour when loaded to the gunwales with CEP2 beta WUs (i.e. when it was still 1:1, not 1:2 as now), but then they're all well balanced systems with enough I/O bandwidth (even the ancient P4 I have at home).

I'm not criticising -- I'm just trying to provide information to throw into the pot for the team to use in whatever way they feel appropriate.

Mike

Edit: Typo.

Just had a check on my betas and in this most recent bunch, 3 have completed in around 8/9 hours (and seem to complete the last 20% or so very quickly, I saw one at around 75% complete then completed about 15mins later), 1 has been completed in exactly 12 hours (I presume this a set cut-off point as this is the second CEP2 beta I've had complete in exactly 12 hours) and the most recent has completed in just over 5 hours.

Seems to be quite a range of completion times on these. They have all started with an estimated completion of around 9 hours then by around 50% seem to be estimated to complete in around 11 hours.

2/3 of the way through my 21 beautiful Beta WU with no errors; 6+ days closer to Ruby
[edited numbers]

Is there any clue as to progress on these CEP2 WU's based on their checkpoint, i.e., does it checkpoint at set points during the run (i.e., at change of seed)?

Edit : After comparing my message log and a WU which I've already sent back, yes, it does look as though each checkpoint is at the end of each job within a work unit. Thus, a clue as to how it's progressing...