Yann Coupin wrote: > Would you mind detailing the process in a more detailed fashion? Load > induced delay shouldn't make the process fail, only take more time. > Failures like this make me think there's a flaw in the design and while > increasing the delay may be a working workaround, it's not completly > errorproof and it slows things down during off-peak hours for no good > reasons.
We are well aware that it's a hack... The basic problem is that the trace ID is allocated by the database, so we have to write the trace record to the database to get the ID before we can move the file to it's final home, as the filename is based on the ID value. The old import daemon wasn't fast enough for it to be a problem but the new one is so we added a delay to make it ignore any traces uploaded in the last few minutes. I came up with a better solution this evening though, so I've just rolled out a change that I hope should fix it. Tom -- Tom Hughes ([email protected]) http://www.compton.nu/ _______________________________________________ talk mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk
