These two bugs are examples where we had bogus locations were enabled in bouncer, for products where Check Now was not set: https://bugzilla.mozilla.org/show_bug.cgi?id=670110 https://bugzilla.mozilla.org/show_bug.cgi?id=668682#c6 Please check what is happening with the sentry run which runs once a day. To flesh out the 2nd case: Was: Product OS Available Thunderbird-3.1.10-Complete win 5000 10 mins after turning on Check Now: Product OS Available Thunderbird-3.1.10-Complete osx 75 Thunderbird-3.1.10-Complete linux 75 Thunderbird-3.1.10-Complete win 75 Thunderbird-3.1.10-Complete solaris-sparc 75 Thunderbird-3.1.10-Complete solaris-i386 75 Thunderbird-3.1.10-Complete opensolaris-sparc 75 Thunderbird-3.1.10-Complete opensolaris-i386 75 So not only did we not have all platforms, we have a mirror which didn't actually have the win files.
So here's the current theory on this: A while back we added a timeout to the sentry script so that if it took over 5 or 6 minutes to check a host (cumulative, across all of the files being checked), it failed out the host assuming it was too slow. There was an exemption to this timeout added to the non-checknow version of the script which runs overnight. The current theory is that this exemption is apparently not working correctly. I haven't gotten a chance to debug it any yet, adding the details in case someone else gets to it before I do.
FWIW, this will be a code issue in the sentry scripts, which are in version control, so anyone could probably poke at it.
Is this still an issue? what's the action needed to close this out? Adding a couple of our coders to cc in case this still needs attention. Please point out where "in version control" they are. thx!
The current issues with sentry's daily run are in bug 716440.