On Mon, 2013-04-15 at 07:18 -0400, Zdenek Pavlas wrote: > > > When there's less mirrors than the retry limit, we should > > > reuse mirrors. A wrapper around the mirror failure callback > > > implements this. > > > Very close to 100% of the time that I move to a different mirror it's > > because I got a 404 from a previous one ... so this is exactly the wrong > > thing to do in that case. > > Unless there's <10 mirrors, it won't make any difference whatsoever.. > > > A significant portion of the rest of the time it's because I got a > > checksum failure, and again this is the wrong thing to do. > > > > Which cases is this going to be good for? > > Single mirror, HTTP 503. Since there's no sleep(), it probably won't > help much, but at least people will stop complaining "there's 10 retries > in yum.conf and it gives up immediately".. > > The failure code IS available in the callback (obj.exception.code), > so we could handle 404 and 503 differently, perhaps not dropping > the mirror only on 503?
Yeh, I'd guess we could say that all 5xx codes are allowed to retry the mirror. ACK with that test. _______________________________________________ Yum-devel mailing list [email protected] http://lists.baseurl.org/mailman/listinfo/yum-devel
