fill server down>broken downloads get stuck

[DeXeS]

Version: 0.3.1
OS: Ubuntu 7.10 Server
Install-type: python source
Template Plush

I had for the last two weeks an extra free news server as "fill server". But the last week, some downloads got stuck when it was a little bit broke (tested that with newsleecher that some files ware incomplete).
the reason why it got stuck is because the free newsserver was down. In the configure screen (plush) the last 10 errors were timeout's to the free fill news server... don't know if it's only me but should be better if when the fill server gets more than 4 timeout's from the fill server, that the fill server is down so it does not try to fetch the broken files from the fill server for eternity.

[alanframe]

Yes I get the same thing. Annoying as the download probably has enough pars to repair but as it's stuck it will never fetch them.

[shypike]

I suppose you are right that a server should not block progress.
Problem is, once a server has been flagged as "down", when do you consider it to be "up" again?

But I am surprised that you actually expect to find missing articles on a free server, while your main server doesn't have them.
This is not exactly my experience.

[DeXeS]

I didn't know that it was down. After that sab started to stop downloading completely, i went investigating and found the fill server down.

And that the freeserver would find the missing articles... yeah, maybe not so clever of me 

better would be if 3 timeouts from fillserver --> stop trying.

[jcfp]

I suppose you are right that a server should not block progress.
Problem is, once a server has been flagged as "down", when do you consider it to be "up" again?

An approach like the one taken by mail servers like postfix might be a good idea: when a server is unavailable, do keep trying, but at increasing intervals (up to a reasonable maximum interval, of course). In the meantime it is flagged as down (thus not blocking anything) and no resources are wasted trying to reconnect in vain yet again every minute or two, while at the same time no manual action or program restart is required to get back to using the server again if/when it eventually comes back.