Hmm, since about 2 weeks I have a really strange thing happened :

When I look at my BOINCstats combined Host Listing, I see many Hosts that have long been replaced by much more modern CPUs.
The Host ID's respective Total Credits and Credits for ongoing Production is okay, however.

Everything was okay until my Primary Project MalariaControl had Problems and I reactivated the Suspended Einstein Accounts on all Systems. The very next day after the Systems had aquired Credit at Einstein, the BOINCstats combined Host Listing turned into what it is now.

Although my Host listings both at Malaria and Einstein show the correct Hosts, BOINCstats combined Host Listing somehow reverted to Hosts that I replaced upto a year ago, while the Individual Host Listings per Project are okay (?)

Direct Comparison by Projects :

MalariaControl, BOINCstats
MalariaControl, Project Stats

Einstein, BOINCstats
Einstein, Project Stats

Looks perfect on the individual Pages (per Project), which puzzles me.

The closest match I can get to the erroneous Overall Host listing is my old Host list from ClimatePrediction (here), but even that is not a 100% match.

------------------
Big Question remaining :
Where does BOINCstats got these obsolete Hosts from and why aren't they replaced by the (correct and current) active Project's Hosts (Malaria, Einstein) ?

Combined hosts are comprised out of the individual project hosts. During the update the projects are merged into one host, and MySQL isn't smart enough to see which CPU is the current one in the system (as a matter of fact: me neither).

Meaning that if a system is upgraded, but doesn't crunch in all projects, there is a good chance that the CPU or OS info is not updated.

Hmm...

Looks then like I will have to find a way to delete my CPDN Hosts somehow, otherwise the Problem will persist indefinitely

The only Stats solution I could think of, would be to flag any Host (field CPU Name) read from a Project XML with 0 RAC (Flag = possibly obsolete) and overwrite it if any other Project has the same Host with >0 RAC (Flag = active).

Another way would be to take the different Host CPU labels taken for the same Host ID and just sort them internally by RAC, and write them to the combined BOINCstats Host list in the proper order (lowest Host ID RAC from any Project is written first into Database, then overwritten by higher RAC from other Project, resulting in the highest Host ID RAC written last and therefor leaving the most current in the combined Stats)
That seems the easiest "failsafe" Workaround.

...possible ?

It's a little more complicated than that. Because of the very large number of hosts the combining of them is not done with a PERL script with lots of options but with a single SQL query. I could set the details of a host with RAC 0 to empty, but of the host is at RAC 0 in all projects it won't have any info left.

I'll let it sink in, one day I'll get an idea on this.

Cool

Would be great to have a solution for this, because otherwise, Users with Obsolete Hosts left in any Project will eventually encounter the very same Problem over time.