[mirror-admin] ERROR: chroot failed for fedora-web
Carlos Carvalho
carlos at fisica.ufpr.br
Fri Jan 16 16:30:59 EST 2009
Paulo Licio de Geus (paulo at las.ic.unicamp.br) wrote on 16 January 2009 10:35:
>Carlos Carvalho wrote:
>
>> Paulo Licio de Geus (paulo at las.ic.unicamp.br) wrote on 15 January 2009 19:41:
>> >Carlos Carvalho wrote:
>> >
>> >> Carlos Carvalho (carlos at fisica.ufpr.br) wrote on 16 January 2009 00:18:
>> >> >I was going to ask you for the specs of report_mirror, that is, what
>> >> >info it sends and how. Then I can provide you the info with what I
>> >> >have here without the dreadful disk scan.
>> >>
>> >> In fact, it'd be much better if we could send only what has changed
>> >> instead of the whole list. Even better, we could just give
>> >> report_mirror a list, either of changes or complete, and it'd transmit
>> >> it as it does now.
>> >>
>> >> --
>> >>
>> >>
>> >
>> >But then out-of-syncs would remain untouched.
>>
>> Not sure what you mean here. Files not mentioned in the changes are
>> supposedly already registered.
>>
>> >A periodic full scan or the whole list available for some more
>> >relaxed, periodic checking would do it...
>>
>> Both the disk scan and sending the full list are necessary only for
>> the first time. That's why I said above that the reporter should
>> accept both changes and the whole list.
>>
>
>Quoting you:
>
>> Exactly. Incremental lists are not robust. Worse, they are even not
>> necessary.
>
>
>Sh*t happens, i.e. failures will occur. The question is when/how your
>out-of-sync server (away from the changed list) will be able to recover.
I said that report_mirror should support both changes-only and full
lists. Mirror updates must be done with the full list always.
--
More information about the Mirror-admin
mailing list