OFFICIAL "SERVER PROBLEMS" THREAD

[petrw1]

Originally Posted by GP2

Does it really need to be run over the entire range up to 1 billion?

It just clears up the active assignments report.

Maybe that report could have a Check Box to only report assignments with
- more than 0% complete
OR
- Activity in the last month or year

[Prime95]

Originally Posted by S485122

I think a query should be run regularly over the whole 0 to 1000M range to eliminate active assignments that are more than 60 (or being generous 90) days late.

IIRC, the query is only run over the active LL and DC assignments range. I don't think it is run on the 100M digit range. I don't remember what it does with TF assignments. I don't have server access right now.

The assignment rules describe when the server has the right to terminate your assignment, the server may not exercise that right.

[S485122]

Originally Posted by Prime95

...
The assignment rules describe when the server has the right to terminate your assignment, the server may not exercise that right.

It is not so much the assignments that are a few days late that worry me, 85% of those lost assignments I found are more than half a year late, 75 % are more than a year late. I don't know what happens in the "active" ranges" since the query I can run is limited to 1000 results, I did not look in the TF ranges for the same reason. Those assignments are preventing others to work on those exponents.

Originally Posted by GP2

...
Consider M332,197,123.
...
Anyone who wanted to could poach it, or maybe a result will miraculously be returned in the meantime.
...

The new rules were put in place to expedite the achievement of milestones, but also to remove the need or the incentive to poach.

Perhaps yearly is often enough to do such a clean up.

Jacob

[VictordeHolland]

Originally Posted by S485122

It is not so much the assignments that are a few days late that worry me, 85% of those lost assignments I found are more than half a year late, 75 % are more than a year late.

As long as I can still get a Manual DoubleCheck assignment and complete it after ~140 days without having to communicate the progress to Primenet (for an offline machine, within the 180 days it says it is valid). I usually pick a high range to not interfere with the milestone progress.

[S485122]

Originally Posted by VictordeHolland

As long as I can still get a Manual DoubleCheck assignment and complete it after ~140 days without having to communicate the progress to Primenet (for an offline machine, within the 180 days it says it is valid). I usually pick a high range to not interfere with the milestone progress.

I don't know how they do it but some assignments have more than 3 years before the next update (perhaps they immediately extend their assignment.)

Jacob

[chalsall]

Originally Posted by S485122

...I did not look in the TF ranges for the same reason. Those assignments are preventing others to work on those exponents.

Just in case you (or anyone else) are concerned about the assignments reserved by GPU72...

We are very careful not to bite off more than we can chew. You can see this in the Estimated Completion report, for example. We only have about 60 days worth of work left to do in the 87M range (to TF up to the optimal level), and it's going to take Primenet's LL'ers about 700 days to bring that to the "wave-back".

In all honesty, what we're currently racing to stay ahead of is the P-1'ers, but even there we're maintaining a comfortable 60 day or so lead.

[Madpoo]

Originally Posted by S485122

I don't know how they do it but some assignments have more than 3 years before the next update (perhaps they immediately extend their assignment.)

Jacob

There are certain exceptions to the expirations, based on the exponent size.

Here are the current exception details as best I can piece together. Basically, these assignments aren't in the way of any milestones or other things and aren't really hurting anyone, so...

First time LL:
Exponents that are 40M larger than the latest LL milestone (currently 78M) but less than 332M (100M digits) are exempted from the typical expiration rules.

First time LL (100M digits between 330M and 350M):
(
If the client has not sent a "next expected" update (basically, they didn't acknowledge the assignment)
*OR* their next update is still a year away)
*OR* there isn't an expected completion date and they have passed their next check-in date
)
*AND* they've only progress less than 2% if it's even started at all

Double checks:
Exponents 20M larger than the latest milestone (currently 43M) are exempted

Smaller PRP (below 40M):
Assignments > 30 days old ... you get 30 days to finish those small PRP checks

Anything that's not DC or LL (TF, P-1, PRP above 40M, etc):
Haven't acknowledged in 1 day since being assigned
*OR*
(
60+ days since the last checkin
*AND* 10+ days since it was last supposed to checkin
)

[S485122]

Originally Posted by chalsall

Just in case you (or anyone else) are concerned about the assignments reserved by GPU72...

I realise that my wording was misleading : the fact that I did not look in certain ranges due to the limits of the queries I can run has nothing to do with holding up some exponents. I meant the two sentences to be unrelated : the forgotten or orphan assignments that I could find are in the way, were I could not look I can't form an opinion :-)

Jacob

[chalsall]

Originally Posted by S485122

...were I could not look I can't form an opinion :-)

Just for the record, I have no privileged access to Primenet's DB nor assignment system.

A trick I myself use when there are more than 1,000 assignments in a range of interest (P-1 assignments in the 85M range, for example) is to first ask for the entire range of interest.

Once this returns the first 1,000 assignments, I then query again with the exponent range starting from the last candidate returned in the previous query. In this example, this is the resulting follow-up query.

Repeat as necessary (or script it).

P.S. To resonate a bit with your premise, it really annoys me when people manually reserve hundreds (sometimes thousands) of P-1 assignments, and then they just sit around for six months before being recycled.

[patrik]

My three cat 1 double checks has just moved into cat 2. Shouldn't that always go the other way?

Also, for double-checks, the thresholds page shows two different exponents for each limit, and has a strange-looking formatting.

[Madpoo]

Originally Posted by patrik

My three cat 1 double checks has just moved into cat 2. Shouldn't that always go the other way?

Also, for double-checks, the thresholds page shows two different exponents for each limit, and has a strange-looking formatting.

I ran into a funny thing yesterday with the daily expirations. I realized I already adjusted the lower bounds for exponent sizes, and forgot that we're not actually up to 43M on the double-checks.

As a result, the 2 exponents below 43M didn't expire and were showing "-1" on the expiration time. Whoops.

So, I fixed it and re-ran just that part of the expiration which probably messed up some formatting, creating double-entries for the thresholds for the day.

Should be all good now.