Testing new AS

[jcoffland]

After a false start we've resumed live testing of the new Assignment Server. This server is a complete rewrite of the old AS. We've been testing it for several months but there were a few bugs that didn't become apparent until we left it in full deployment for a couple of days. We have identified and resolved several issues but there may be more. Please be patient while we get this straightened out. We will fall back to the old AS again if more problems occur.

The new AS has a lot of features that will help us detect problems earlier and provide us with better tools for analyzing problems when they do occur. So there may be a few more bumps but you can expect better resource allocation and less problems in the future.

Thanks,

Joseph

[jcoffland]

FYI, I've rolled back to the old AS for now since we don't like to run new code over the weekend. I will redeploy the new AS on Monday.

[jcoffland]

The new AS is back in action. I believe I fixed the problems we saw last week but please report any assignment errors.

The new AS gives us a much better picture of what's going on in the F@H network. I see that we are low on jobs for older Nvidia GPUs and uniprocessor clients.

[bruce]

According to the information I have received, at least three problems have been fixed today.

1) Maxwell was not getting Core_17. Fixed based on the availability of Core_18 assignments or on Core_17 in Advanced. UPGRADE TO THE LATEST DRIVERS.
2) Clients that are restricted to port 80 have had problems getting 0x17 assignments if client-type=FAH (default) or =ADVANCED. Fixed.
3) V6 SMP and V7 seem to have been working but there was a problem with v6 GPU clients. Fixed.

Going forward, please avoid additional discussion unless
a) those issues have not been resolved or
b) you're reporting a different problem.