Trial Factor and P-1 assignments squatting in Cat 0 and Cat 1 ranges

[chalsall]

Originally Posted by Uncwilly

ATH and S485122 have suggested server rules that would prevent this.

Let me please suggest another...

TF assignments are recycled after eight (8#) days without an update. Regardless of Anonymous or signed-in user.

[kriesel]

Originally Posted by chalsall

Let me please suggest another...

TF assignments are recycled after eight (8#) days without an update. Regardless of Anonymous or signed-in user.

Not a fan of an 8-day deadline. I maintain a minimum queue depth of 2 days work per gpu process (if the gpu had nothing else to do, no other processes running). On gtx1650 that's 2 processes, RTX2080x, 3 processes. Either gpu model will routinely go above 8 days total queued work per gpu, since I replenish each process with a 20-pack or 30-pack of Factor=x,whatever,76. That's for the production stuff intended to sweep TF for the first-test wavefront.

At high exponent runs such as I have going currently, they may be going to anywhere from 77 to 86 bits and an individual assignment may take weeks.
It's all manual, so there are no updates.

[chalsall]

Originally Posted by kriesel

It's all manual, so there are no updates.

Sucks to be you...

What about 30 days? The current 180 is ridiculous!

Also, of course, the other suggestions above... Don't allow assignments that don't make sense seems like a pretty easy one to implement.

People are always free to do work which tickles their fancy "off the books". A thread somewhere here on the Forum (or even just PMs) can avoid toe compression...

[LaurV]

Originally Posted by chalsall

Sucks to be you...

Uncalled for! I like you more than Ken, but I am with him here

I am against 8 days too. I had issues with my systems (broken pump, fan, etc) and not having immediate time to fix the issue, or occasionally went out of the city for holiday or job related, when the wheelbarrows were not crunching for days, but I NEVER let my assignments unfinished. And yes, they were automatically fetched by Misfit, so, "don't take so many assignments then", doesn't work.

30 days is kinda borderline of what would be acceptable.

[Viliam Furik]

I expect it will take up to a week to complete my 158 assignments in the FTC range.

If it's that much of a problem, I won't object to dropping them.

Would off-the-books TF-ing help the situation? Factoring Limits site can give work without assigning it officially.

[ATH]

I'm not sure why you keep doing TF in those ranges when we keep making these threads complaining about those TF assignments? TF is not needed in Cat0+1+2 ranges.

[retina]

Originally Posted by Viliam Furik

Would off-the-books TF-ing help the situation?

No.

If you really want to help, do TF in another range, or do PRP.

[Viliam Furik]

Originally Posted by retina

No.

If you really want to help, do TF in another range, or do PRP.

Ok, noted. But I meant help with the blocking of assignments.

---------------
EDIT: They are dropped.

[chalsall]

Originally Posted by Viliam Furik

Ok, noted. But I meant help with the blocking of assignments. EDIT: They are dropped.

To answer your question...

Yes... Doing "past optimal" TF'ing work "off-the-books" would prevent the issue we're complaining about here. Specifically, it will avoid the situation where a FC (or DC) assignment can't be given because someone is holding it for TF'ing.

There is the argument this shouldn't be done, because a factor could be found for a candidate that is already undergoing a FC/DC.

But in my mind this isn't a huge issue. Appoximately a 1% chance that will occure, and it's not any different than someone doing further TF'ing on a candidate already worked.

[retina]

Originally Posted by chalsall

There is the argument this shouldn't be done, because a factor could be found for a candidate that is already undergoing a FC/DC.

That's not the reason to not do it. By doing unneeded TF in one range (regardless of off- or on-book), other more needed ranges are being ignored.

[Mark Rose]

Originally Posted by LaurV

Uncalled for! I like you more than Ken, but I am with him here

I am against 8 days too. I had issues with my systems (broken pump, fan, etc) and not having immediate time to fix the issue, or occasionally went out of the city for holiday or job related, when the wheelbarrows were not crunching for days, but I NEVER let my assignments unfinished. And yes, they were automatically fetched by Misfit, so, "don't take so many assignments then", doesn't work.

30 days is kinda borderline of what would be acceptable.

This happened to me. Just last night I reported 3+ weeks of work on a system. It was a system I freshly reinstalled before going on vacation, I had manually given it a pile of assignments, which all completed, but the fetching/submission agent wasn't working correctly. That was about 14,000 GHz-days of work.

30 days would be the minimum I'd be happy with, for situations like this, but I agree 180 is far too much.