In my section of the report that says:



This is drawn directly from the database (i.e. not me speculating).

However, a specific workunit only gets marked as being "Extreme", "Accelerated" or "Normal" while it is transitioning from one generation to the next. Thus if we generation 105 beomes the " Accelerated" generation, then workunits that have already been created as generation 105 will not suddenly be marked as "Accelerated". Only those workunits that transition into that generation (or less) get marked as "Accelerated". Thus you can receive jobs from a generation that is now the "Accelerated" generation that are not themselves "Accelerated" because they were created before that generation became "Accelerated".

An additional comment, is that I use my report for tracking progress on a daily basis - so I have written the SQL query in a way that it gives me the data as it exists at 12:00 UTC on the current date.

Thank you, Kevin.

That explains the situation with regard to generation 087/088. I will ignore the latest generation in my stats. as many of them will have been issued before the promotion.

The same will go for the 103/104, but it is pushing the limits, but not impossible, to explain why 104's were still being issued without priority some 18 hours later. The cache at WCG must be higher than previously.

I will adjust my end date forecast by half a day to allow for the noon dump instead of midnight.

Kind regards

Mike

Ok - there was an issue with my query and it did falsely report 104 has been accelerated. Here is the corrected data:

number_units max_generation type        
------------ -------------- ----------- 
31           088            Extreme     
10628        103            Accelerated 
24950        112            Normal      

Mark

From my observations, resends do not get priority status unless they are already past their original half-way point.

Mike

entity and Mike,

Thank you for taking the time to explain this to me. As a newbie cruncher, I've got a lot to learn!

The examples I noticed of normal generation resends with the standard deadline seem to have occurred in cases where one of the original tasks errored out more or less immediately, so that would be consistent with Mike's observation.

Cheers,
Mark