Unixchick

A problem like the current one does not show up immediately on the statistics. They are still sending out work, but only resends, so many people are still working at 100%. I hold 5 units at any one time of which 1 is a spare. I have been getting mostly resends anyway. My spare was utilised and took some time to get replaced, but it did get replaced before another would have been needed. This happened twice today, but I did not lose any crunching time.

Resends will become scarcer and then we will get a significant fall in results returned, probably today.

Mike.

I am wondering about your previous observation about the validations. Do you think that means there are more crunchers than they can handle?

I am happy to contribute if they need me, but there are a lot of projects that really don't have enough work. The symptoms may be different, but the occasional outages here may be an indication of that. Your insight would help on that.

Jim1348

Once this problem is sorted, they will need you.

What is needed is for certain crunchers not to hoard units. It slows down the validation and therefore the creation of the next generation.

I would suggest that no-one should hold more than one spare unit and anyone taking more than 2 days per unit should not hold any spares.

As a general rule, the faster machines have more threads available.

A 16 thread machine running ARP on 8 threads (and other projects on the other threads) will probably take about 8-12 hours for each unit, so the spare unit would be spare for only about 60-90 minutes.

An 8 thread machine running ARP on 4 threads will probably take about 24 hours so the spare would be spare for about 6 hours.

Those times for the spare are acceptable, but a machine taking over 2 days to crunch a unit would probably have a spare hanging around for a day or more which would not be desirable. Apart from delaying the process, it would mean that the spare would cause the machine not to be classified as 'reliable' and so not get priority generations or resends.

There are only 35,609 units available so 71,218 including the wingmen, but that number is reduced by the number of results awaiting validation, hence the need to minimise the spares.

Mike

It takes quite a while owing to the hoarded units and there is always a proportion of resends due to 'no replies'. That group takes 8 days to become resends.

Without a fix, there would probably be a few stragglers still being crunched next week.

Mike

I thought these tasks would all be stragglers, but the deadlines differs:

ARP1_0019642_073_1 - 01 Oct 04:41:54
ARP1_0014175_074_0 - 01 Oct 04:41:54
ARP1_0028643_076_1 - 04 Oct 11:30:36
ARP1_0017718_077_0 - 01 Oct 04:41:54
ARP1_0012938_082_1 - 04 Oct 11:30:36
ARP1_0004519_084_0 - 01 Oct 04:41:54
ARP1_0029315_086_0 - 03 Oct 00:01:02