Re: Bug#465434: qa.debian.org: Uscan Errors processing Watch File
Bart Martens wrote:
> On Sun, May 17, 2009 at 11:27:25AM -0500, Raphael Geissert wrote:
[...]
>
> Can you modify the rescanning so that such 403's are
> rescanned sooner than with the next total rescan every 4 days ? Note that
No matter how good that idea may sound, it isn't. By re-checking them all in
one go they are more prone to be checked together once again, and if a
temporary error affects them, they will all be affected.
There used to be a weekly cronjob that re-checked all the failed watch
files, but it didn't work well for the above mentioned problem.
> such 403's may (I'm guessing) repeat to be 403's as long as they're in the
> total rescan.
There's no such total rescan; rechecking all the 8431 watch files every four
days in one go would degrade the performance of the system. Every watch
file is independent of each other (and so their last-checked timestamp).
> Maybe
> semi-automatically/manually rescan selectively and inject the successful
> results at some convenient point in the data flow to the
> qa-developer-pages ? It could be an effectively working workaround until
> DEHS2 is released.
>
The BDBs and all the files used by external systems are regenerated every
time a run finishes (that includes the daily new packages processing and
the daily general processing).
>>
>> >> > Displaying the old version on the
>> >> > qa-developer-pages until the next rescan is successful.
>> >>
>> >> Something similar is planned for DEHS2
>> >
>> > Release in 2009 ?
>>
>> Of course; but some major work needs to be done in uscan first.
>
> I understand that major work may be required for DEHS2, but I don't think
> that uscan needs any changes to have the old rescan results displayed on
> the
> qa-developer-pages until the next rescan is successful.
No, they aren't; but for uscan to be useful to DEHS2 it will be turned into
a perl module. No more workarounds for DEHS are going to be made, unless
there's a critical issue that needs to be fixed (and I don't expect any,
since the same code base has been used since 2004).
> Any chance that
> this gets planned a bit earlier, so before the entire DEHS2 ?
>
Nope
>> >> > Re-scanning selectively briefly (minutes/hours) after the failed
>> >> > scans to deal with temporary errors.
>> >>
>> >> No, that won't happen as it would be easier to get alioth banned.
>> >
>> > Isn't waiting an hour or 2 and try again exactly the approach to work
>> > around such temporary ban ? Or do you fear a permanent ban ?
>> >
>>
>> I'm not fearing a permanent ban, but for example berlios' ban lasts a
>> day.
>
> So for Berlios re-scanning selectively briefly after the failed scans to
> deal with temporary errors could be done after one day, obviously.
That's exactly what happens, every four days.
Regards,
--
Raphael Geissert - Debian Maintainer
www.debian.org - get.debian.net
Reply to: