Andre Hinrichs <[EMAIL PROTECTED]> writes: > On Mittwoch, 19. November 2008, Dirk-Lüder Kreie wrote: >> it's far better to use loop mode instead of a batch mode, especially >> when interacting with the tah server, for a number of reasons: >> the loop mode: >> a) handles uploading in the background automagically > > For which I don't care if I'm running a dedicated upload_loop.
As spaetz pointed out the idea is that all clients upload to the dedicated uploader which needs to be a separate installation because it needs a different config. Of course, since this is just a file copy there is not much benefit in doing this in the background. It just occured to me that this is far from optimal. Instead, a client running in upload_loop mode should simply upload from the UploadDirectory. This would eliminate the need for a separate setup. > >> b) exits if there is a persistent fault (some external tool broken) > > That's why I wrote the wrapper script. The client stopped on rare cases of > errors. But you are right that letting a client restart even if there are > persistent errors would be a bad thing. Question is how to distinguish > between rare errors and persistent errors. There are some tiles which cause > errors here (the libGD error) and that's why I asked in another mail if there > is a possibility to put a job back manually so that the community do not have > to wait for the seven hours lasting timeout. And I didn't want to let this > rare cases stop the loop. The approach the client currently takes is to count errors of certain types and to quit if an error occurs more than a certain number of times in a row. Unfortunately, this has been neglected lately and needs some overhaul. This is why some errors are handled in the wrong way. Matthias _______________________________________________ Tilesathome mailing list [email protected] http://lists.openstreetmap.org/listinfo/tilesathome
