Drop block that sets max mirror tries to 1

[dcantrell]

This functionality has always been present, but it's not clear why. It was never requested by any user. Removing these lines does allow everything to function as expected, but dnf makecache now retries when mirrors fail.

Resolves: RHEL-1342

[m-blaha]

FTR - it was introduced by commit d48bc7c as a fix for bug https://bugzilla.redhat.com/show_bug.cgi?id=922667

[jan-kolarik]

FTR - it was introduced by commit d48bc7c as a fix for bug https://bugzilla.redhat.com/show_bug.cgi?id=922667

Indeed, Marek, you're right. I've traced back through the history up to this PR, which appeared to introduce this functionality as the initial version, but I didn't notice the refactoring there 🙁

Now, it appears we have two somewhat conflicting requests. The older one suggests trying only the first mirror and skipping the others if it fails, while the new one proposes attempting a different one specifically in the case of a certificate problem. However, this distinction isn't recognized by librepo, which only has the LRO_MAXMIRRORTRIES option.

If we follow the suggestion in #1999 and remove the --timer parameter from the cron job, it will address the newer request related to certificate problems. Still, the older one won't be satisfied, and we will break the resource-awareness of the job, as reported in the past here. However, it's uncertain how relevant these problems are in the current context.

[jan-kolarik]

@dcantrell Do you have any ideas how to proceed? Maybe we should bring this up for discussion during the subsystem meeting...