Between alpha and beta freezes, I decreased the MM config value max_stale_days from 7 to 2. This causes MM to "forget" about old repomd.xml data after 2 days instead of 7, which reduces how long a mirror can be "stale" before being ignored by yum. As we've slowed down releasing updates for existing releases, it can be longer than 2 days between pushing updates. In this case, we run into the propogation delay between when content gets posted on the master, MM picking it up an hour later, and then "drops" the now-historical metadata that's older than 2 days - mirrors who haven't updated within that hour get ignored by yum with a pretty loud error message. I've got a fix, but await end of change freeze to roll out a new MM with the fix. Until then, I'd like to return the config value back to 7 days, which papers over the propogation delay much better. +1s please. Thanks, Matt -- Matt Domsch Technology Strategist Dell | Office of the CTO _______________________________________________ infrastructure mailing list infrastructure@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/infrastructure
